FreeBSD/Linux Kernel Cross Reference
sys/netinet/ip_fw2.c
1 /*
2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
6 * are met:
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
12 *
13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23 * SUCH DAMAGE.
24 *
25 * $FreeBSD: releng/5.3/sys/netinet/ip_fw2.c 136588 2004-10-16 08:43:07Z cvs2svn $
26 */
27
28 #define DEB(x)
29 #define DDB(x) x
30
31 /*
32 * Implement IP packet firewall (new version)
33 */
34
35 #if !defined(KLD_MODULE)
36 #include "opt_ipfw.h"
37 #include "opt_ipdn.h"
38 #include "opt_ipdivert.h"
39 #include "opt_inet.h"
40 #include "opt_ipsec.h"
41 #ifndef INET
42 #error IPFIREWALL requires INET.
43 #endif /* INET */
44 #endif
45
46 #define IPFW2 1
47 #if IPFW2
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/kernel.h>
53 #include <sys/jail.h>
54 #include <sys/module.h>
55 #include <sys/proc.h>
56 #include <sys/socket.h>
57 #include <sys/socketvar.h>
58 #include <sys/sysctl.h>
59 #include <sys/syslog.h>
60 #include <sys/ucred.h>
61 #include <net/if.h>
62 #include <net/radix.h>
63 #include <net/route.h>
64 #include <netinet/in.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/in_var.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/ip.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/ip_icmp.h>
71 #include <netinet/ip_fw.h>
72 #include <netinet/ip_divert.h>
73 #include <netinet/ip_dummynet.h>
74 #include <netinet/tcp.h>
75 #include <netinet/tcp_timer.h>
76 #include <netinet/tcp_var.h>
77 #include <netinet/tcpip.h>
78 #include <netinet/udp.h>
79 #include <netinet/udp_var.h>
80
81 #ifdef IPSEC
82 #include <netinet6/ipsec.h>
83 #endif
84
85 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
86
87 #include <machine/in_cksum.h> /* XXX for in_cksum */
88
89 /*
90 * set_disable contains one bit per set value (0..31).
91 * If the bit is set, all rules with the corresponding set
92 * are disabled. Set RESVD_SET(31) is reserved for the default rule
93 * and rules that are not deleted by the flush command,
94 * and CANNOT be disabled.
95 * Rules in set RESVD_SET can only be deleted explicitly.
96 */
97 static u_int32_t set_disable;
98
99 static int fw_verbose;
100 static int verbose_limit;
101
102 static struct callout ipfw_timeout;
103 #define IPFW_DEFAULT_RULE 65535
104
105 /*
106 * Data structure to cache our ucred related
107 * information. This structure only gets used if
108 * the user specified UID/GID based constraints in
109 * a firewall rule.
110 */
111 struct ip_fw_ugid {
112 gid_t fw_groups[NGROUPS];
113 int fw_ngroups;
114 uid_t fw_uid;
115 int fw_prid;
116 };
117
118 struct ip_fw_chain {
119 struct ip_fw *rules; /* list of rules */
120 struct ip_fw *reap; /* list of rules to reap */
121 struct mtx mtx; /* lock guarding rule list */
122 };
123 #define IPFW_LOCK_INIT(_chain) \
124 mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \
125 MTX_DEF | MTX_RECURSE)
126 #define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx)
127 #define IPFW_LOCK(_chain) mtx_lock(&(_chain)->mtx)
128 #define IPFW_UNLOCK(_chain) mtx_unlock(&(_chain)->mtx)
129 #define IPFW_LOCK_ASSERT(_chain) do { \
130 mtx_assert(&(_chain)->mtx, MA_OWNED); \
131 NET_ASSERT_GIANT(); \
132 } while (0)
133
134 /*
135 * list of rules for layer 3
136 */
137 static struct ip_fw_chain layer3_chain;
138
139 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
140 MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables");
141
142 struct table_entry {
143 struct radix_node rn[2];
144 struct sockaddr_in addr, mask;
145 u_int32_t value;
146 };
147
148 #define IPFW_TABLES_MAX 128
149 static struct {
150 struct radix_node_head *rnh;
151 int modified;
152 } ipfw_tables[IPFW_TABLES_MAX];
153
154 static int fw_debug = 1;
155 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
156
157 #ifdef SYSCTL_NODE
158 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
159 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable,
160 CTLFLAG_RW | CTLFLAG_SECURE3,
161 &fw_enable, 0, "Enable ipfw");
162 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
163 &autoinc_step, 0, "Rule number autincrement step");
164 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
165 CTLFLAG_RW | CTLFLAG_SECURE3,
166 &fw_one_pass, 0,
167 "Only do a single pass through ipfw when using dummynet(4)");
168 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW,
169 &fw_debug, 0, "Enable printing of debug ip_fw statements");
170 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
171 CTLFLAG_RW | CTLFLAG_SECURE3,
172 &fw_verbose, 0, "Log matches to ipfw rules");
173 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
174 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
175
176 /*
177 * Description of dynamic rules.
178 *
179 * Dynamic rules are stored in lists accessed through a hash table
180 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
181 * be modified through the sysctl variable dyn_buckets which is
182 * updated when the table becomes empty.
183 *
184 * XXX currently there is only one list, ipfw_dyn.
185 *
186 * When a packet is received, its address fields are first masked
187 * with the mask defined for the rule, then hashed, then matched
188 * against the entries in the corresponding list.
189 * Dynamic rules can be used for different purposes:
190 * + stateful rules;
191 * + enforcing limits on the number of sessions;
192 * + in-kernel NAT (not implemented yet)
193 *
194 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
195 * measured in seconds and depending on the flags.
196 *
197 * The total number of dynamic rules is stored in dyn_count.
198 * The max number of dynamic rules is dyn_max. When we reach
199 * the maximum number of rules we do not create anymore. This is
200 * done to avoid consuming too much memory, but also too much
201 * time when searching on each packet (ideally, we should try instead
202 * to put a limit on the length of the list on each bucket...).
203 *
204 * Each dynamic rule holds a pointer to the parent ipfw rule so
205 * we know what action to perform. Dynamic rules are removed when
206 * the parent rule is deleted. XXX we should make them survive.
207 *
208 * There are some limitations with dynamic rules -- we do not
209 * obey the 'randomized match', and we do not do multiple
210 * passes through the firewall. XXX check the latter!!!
211 */
212 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
213 static u_int32_t dyn_buckets = 256; /* must be power of 2 */
214 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */
215
216 static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */
217 #define IPFW_DYN_LOCK_INIT() \
218 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF)
219 #define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx)
220 #define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx)
221 #define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx)
222 #define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED)
223
224 /*
225 * Timeouts for various events in handing dynamic rules.
226 */
227 static u_int32_t dyn_ack_lifetime = 300;
228 static u_int32_t dyn_syn_lifetime = 20;
229 static u_int32_t dyn_fin_lifetime = 1;
230 static u_int32_t dyn_rst_lifetime = 1;
231 static u_int32_t dyn_udp_lifetime = 10;
232 static u_int32_t dyn_short_lifetime = 5;
233
234 /*
235 * Keepalives are sent if dyn_keepalive is set. They are sent every
236 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
237 * seconds of lifetime of a rule.
238 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
239 * than dyn_keepalive_period.
240 */
241
242 static u_int32_t dyn_keepalive_interval = 20;
243 static u_int32_t dyn_keepalive_period = 5;
244 static u_int32_t dyn_keepalive = 1; /* do send keepalives */
245
246 static u_int32_t static_count; /* # of static rules */
247 static u_int32_t static_len; /* size in bytes of static rules */
248 static u_int32_t dyn_count; /* # of dynamic rules */
249 static u_int32_t dyn_max = 4096; /* max # of dynamic rules */
250
251 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
252 &dyn_buckets, 0, "Number of dyn. buckets");
253 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
254 &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
255 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
256 &dyn_count, 0, "Number of dyn. rules");
257 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
258 &dyn_max, 0, "Max number of dyn. rules");
259 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
260 &static_count, 0, "Number of static rules");
261 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
262 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
263 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
264 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
265 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
266 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
267 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
268 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
269 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
270 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
271 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
272 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
273 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
274 &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
275
276 #endif /* SYSCTL_NODE */
277
278
279 /*
280 * This macro maps an ip pointer into a layer3 header pointer of type T
281 */
282 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
283
284 static __inline int
285 icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd)
286 {
287 int type = L3HDR(struct icmp,ip)->icmp_type;
288
289 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
290 }
291
292 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
293 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
294
295 static int
296 is_icmp_query(struct ip *ip)
297 {
298 int type = L3HDR(struct icmp, ip)->icmp_type;
299 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
300 }
301 #undef TT
302
303 /*
304 * The following checks use two arrays of 8 or 16 bits to store the
305 * bits that we want set or clear, respectively. They are in the
306 * low and high half of cmd->arg1 or cmd->d[0].
307 *
308 * We scan options and store the bits we find set. We succeed if
309 *
310 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
311 *
312 * The code is sometimes optimized not to store additional variables.
313 */
314
315 static int
316 flags_match(ipfw_insn *cmd, u_int8_t bits)
317 {
318 u_char want_clear;
319 bits = ~bits;
320
321 if ( ((cmd->arg1 & 0xff) & bits) != 0)
322 return 0; /* some bits we want set were clear */
323 want_clear = (cmd->arg1 >> 8) & 0xff;
324 if ( (want_clear & bits) != want_clear)
325 return 0; /* some bits we want clear were set */
326 return 1;
327 }
328
329 static int
330 ipopts_match(struct ip *ip, ipfw_insn *cmd)
331 {
332 int optlen, bits = 0;
333 u_char *cp = (u_char *)(ip + 1);
334 int x = (ip->ip_hl << 2) - sizeof (struct ip);
335
336 for (; x > 0; x -= optlen, cp += optlen) {
337 int opt = cp[IPOPT_OPTVAL];
338
339 if (opt == IPOPT_EOL)
340 break;
341 if (opt == IPOPT_NOP)
342 optlen = 1;
343 else {
344 optlen = cp[IPOPT_OLEN];
345 if (optlen <= 0 || optlen > x)
346 return 0; /* invalid or truncated */
347 }
348 switch (opt) {
349
350 default:
351 break;
352
353 case IPOPT_LSRR:
354 bits |= IP_FW_IPOPT_LSRR;
355 break;
356
357 case IPOPT_SSRR:
358 bits |= IP_FW_IPOPT_SSRR;
359 break;
360
361 case IPOPT_RR:
362 bits |= IP_FW_IPOPT_RR;
363 break;
364
365 case IPOPT_TS:
366 bits |= IP_FW_IPOPT_TS;
367 break;
368 }
369 }
370 return (flags_match(cmd, bits));
371 }
372
373 static int
374 tcpopts_match(struct ip *ip, ipfw_insn *cmd)
375 {
376 int optlen, bits = 0;
377 struct tcphdr *tcp = L3HDR(struct tcphdr,ip);
378 u_char *cp = (u_char *)(tcp + 1);
379 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
380
381 for (; x > 0; x -= optlen, cp += optlen) {
382 int opt = cp[0];
383 if (opt == TCPOPT_EOL)
384 break;
385 if (opt == TCPOPT_NOP)
386 optlen = 1;
387 else {
388 optlen = cp[1];
389 if (optlen <= 0)
390 break;
391 }
392
393 switch (opt) {
394
395 default:
396 break;
397
398 case TCPOPT_MAXSEG:
399 bits |= IP_FW_TCPOPT_MSS;
400 break;
401
402 case TCPOPT_WINDOW:
403 bits |= IP_FW_TCPOPT_WINDOW;
404 break;
405
406 case TCPOPT_SACK_PERMITTED:
407 case TCPOPT_SACK:
408 bits |= IP_FW_TCPOPT_SACK;
409 break;
410
411 case TCPOPT_TIMESTAMP:
412 bits |= IP_FW_TCPOPT_TS;
413 break;
414
415 case TCPOPT_CC:
416 case TCPOPT_CCNEW:
417 case TCPOPT_CCECHO:
418 bits |= IP_FW_TCPOPT_CC;
419 break;
420 }
421 }
422 return (flags_match(cmd, bits));
423 }
424
425 static int
426 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
427 {
428 if (ifp == NULL) /* no iface with this packet, match fails */
429 return 0;
430 /* Check by name or by IP address */
431 if (cmd->name[0] != '\0') { /* match by name */
432 /* Check name */
433 if (cmd->p.glob) {
434 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
435 return(1);
436 } else {
437 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
438 return(1);
439 }
440 } else {
441 struct ifaddr *ia;
442
443 /* XXX lock? */
444 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
445 if (ia->ifa_addr == NULL)
446 continue;
447 if (ia->ifa_addr->sa_family != AF_INET)
448 continue;
449 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
450 (ia->ifa_addr))->sin_addr.s_addr)
451 return(1); /* match */
452 }
453 }
454 return(0); /* no match, fail ... */
455 }
456
457 /*
458 * The verify_path function checks if a route to the src exists and
459 * if it is reachable via ifp (when provided).
460 *
461 * The 'verrevpath' option checks that the interface that an IP packet
462 * arrives on is the same interface that traffic destined for the
463 * packet's source address would be routed out of. The 'versrcreach'
464 * option just checks that the source address is reachable via any route
465 * (except default) in the routing table. These two are a measure to block
466 * forged packets. This is also commonly known as "anti-spoofing" or Unicast
467 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
468 * is purposely reminiscent of the Cisco IOS command,
469 *
470 * ip verify unicast reverse-path
471 * ip verify unicast source reachable-via any
472 *
473 * which implements the same functionality. But note that syntax is
474 * misleading. The check may be performed on all IP packets whether unicast,
475 * multicast, or broadcast.
476 */
477 static int
478 verify_path(struct in_addr src, struct ifnet *ifp)
479 {
480 struct route ro;
481 struct sockaddr_in *dst;
482
483 bzero(&ro, sizeof(ro));
484
485 dst = (struct sockaddr_in *)&(ro.ro_dst);
486 dst->sin_family = AF_INET;
487 dst->sin_len = sizeof(*dst);
488 dst->sin_addr = src;
489 rtalloc_ign(&ro, RTF_CLONING);
490
491 if (ro.ro_rt == NULL)
492 return 0;
493
494 /* if ifp is provided, check for equality with rtentry */
495 if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) {
496 RTFREE(ro.ro_rt);
497 return 0;
498 }
499
500 /* if no ifp provided, check if rtentry is not default route */
501 if (ifp == NULL &&
502 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
503 RTFREE(ro.ro_rt);
504 return 0;
505 }
506
507 /* or if this is a blackhole/reject route */
508 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
509 RTFREE(ro.ro_rt);
510 return 0;
511 }
512
513 /* found valid route */
514 RTFREE(ro.ro_rt);
515 return 1;
516 }
517
518
519 static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */
520
521 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
522 #define SNP(buf) buf, sizeof(buf)
523
524 /*
525 * We enter here when we have a rule with O_LOG.
526 * XXX this function alone takes about 2Kbytes of code!
527 */
528 static void
529 ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh,
530 struct mbuf *m, struct ifnet *oif)
531 {
532 char *action;
533 int limit_reached = 0;
534 char action2[40], proto[48], fragment[28];
535
536 fragment[0] = '\0';
537 proto[0] = '\0';
538
539 if (f == NULL) { /* bogus pkt */
540 if (verbose_limit != 0 && norule_counter >= verbose_limit)
541 return;
542 norule_counter++;
543 if (norule_counter == verbose_limit)
544 limit_reached = verbose_limit;
545 action = "Refuse";
546 } else { /* O_LOG is the first action, find the real one */
547 ipfw_insn *cmd = ACTION_PTR(f);
548 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
549
550 if (l->max_log != 0 && l->log_left == 0)
551 return;
552 l->log_left--;
553 if (l->log_left == 0)
554 limit_reached = l->max_log;
555 cmd += F_LEN(cmd); /* point to first action */
556 if (cmd->opcode == O_PROB)
557 cmd += F_LEN(cmd);
558
559 action = action2;
560 switch (cmd->opcode) {
561 case O_DENY:
562 action = "Deny";
563 break;
564
565 case O_REJECT:
566 if (cmd->arg1==ICMP_REJECT_RST)
567 action = "Reset";
568 else if (cmd->arg1==ICMP_UNREACH_HOST)
569 action = "Reject";
570 else
571 snprintf(SNPARGS(action2, 0), "Unreach %d",
572 cmd->arg1);
573 break;
574
575 case O_ACCEPT:
576 action = "Accept";
577 break;
578 case O_COUNT:
579 action = "Count";
580 break;
581 case O_DIVERT:
582 snprintf(SNPARGS(action2, 0), "Divert %d",
583 cmd->arg1);
584 break;
585 case O_TEE:
586 snprintf(SNPARGS(action2, 0), "Tee %d",
587 cmd->arg1);
588 break;
589 case O_SKIPTO:
590 snprintf(SNPARGS(action2, 0), "SkipTo %d",
591 cmd->arg1);
592 break;
593 case O_PIPE:
594 snprintf(SNPARGS(action2, 0), "Pipe %d",
595 cmd->arg1);
596 break;
597 case O_QUEUE:
598 snprintf(SNPARGS(action2, 0), "Queue %d",
599 cmd->arg1);
600 break;
601 case O_FORWARD_IP: {
602 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
603 int len;
604
605 len = snprintf(SNPARGS(action2, 0), "Forward to %s",
606 inet_ntoa(sa->sa.sin_addr));
607 if (sa->sa.sin_port)
608 snprintf(SNPARGS(action2, len), ":%d",
609 sa->sa.sin_port);
610 }
611 break;
612 default:
613 action = "UNKNOWN";
614 break;
615 }
616 }
617
618 if (hlen == 0) { /* non-ip */
619 snprintf(SNPARGS(proto, 0), "MAC");
620 } else {
621 struct ip *ip = mtod(m, struct ip *);
622 /* these three are all aliases to the same thing */
623 struct icmp *const icmp = L3HDR(struct icmp, ip);
624 struct tcphdr *const tcp = (struct tcphdr *)icmp;
625 struct udphdr *const udp = (struct udphdr *)icmp;
626
627 int ip_off, offset, ip_len;
628
629 int len;
630
631 if (eh != NULL) { /* layer 2 packets are as on the wire */
632 ip_off = ntohs(ip->ip_off);
633 ip_len = ntohs(ip->ip_len);
634 } else {
635 ip_off = ip->ip_off;
636 ip_len = ip->ip_len;
637 }
638 offset = ip_off & IP_OFFMASK;
639 switch (ip->ip_p) {
640 case IPPROTO_TCP:
641 len = snprintf(SNPARGS(proto, 0), "TCP %s",
642 inet_ntoa(ip->ip_src));
643 if (offset == 0)
644 snprintf(SNPARGS(proto, len), ":%d %s:%d",
645 ntohs(tcp->th_sport),
646 inet_ntoa(ip->ip_dst),
647 ntohs(tcp->th_dport));
648 else
649 snprintf(SNPARGS(proto, len), " %s",
650 inet_ntoa(ip->ip_dst));
651 break;
652
653 case IPPROTO_UDP:
654 len = snprintf(SNPARGS(proto, 0), "UDP %s",
655 inet_ntoa(ip->ip_src));
656 if (offset == 0)
657 snprintf(SNPARGS(proto, len), ":%d %s:%d",
658 ntohs(udp->uh_sport),
659 inet_ntoa(ip->ip_dst),
660 ntohs(udp->uh_dport));
661 else
662 snprintf(SNPARGS(proto, len), " %s",
663 inet_ntoa(ip->ip_dst));
664 break;
665
666 case IPPROTO_ICMP:
667 if (offset == 0)
668 len = snprintf(SNPARGS(proto, 0),
669 "ICMP:%u.%u ",
670 icmp->icmp_type, icmp->icmp_code);
671 else
672 len = snprintf(SNPARGS(proto, 0), "ICMP ");
673 len += snprintf(SNPARGS(proto, len), "%s",
674 inet_ntoa(ip->ip_src));
675 snprintf(SNPARGS(proto, len), " %s",
676 inet_ntoa(ip->ip_dst));
677 break;
678
679 default:
680 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p,
681 inet_ntoa(ip->ip_src));
682 snprintf(SNPARGS(proto, len), " %s",
683 inet_ntoa(ip->ip_dst));
684 break;
685 }
686
687 if (ip_off & (IP_MF | IP_OFFMASK))
688 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)",
689 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
690 offset << 3,
691 (ip_off & IP_MF) ? "+" : "");
692 }
693 if (oif || m->m_pkthdr.rcvif)
694 log(LOG_SECURITY | LOG_INFO,
695 "ipfw: %d %s %s %s via %s%s\n",
696 f ? f->rulenum : -1,
697 action, proto, oif ? "out" : "in",
698 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname,
699 fragment);
700 else
701 log(LOG_SECURITY | LOG_INFO,
702 "ipfw: %d %s %s [no if info]%s\n",
703 f ? f->rulenum : -1,
704 action, proto, fragment);
705 if (limit_reached)
706 log(LOG_SECURITY | LOG_NOTICE,
707 "ipfw: limit %d reached on entry %d\n",
708 limit_reached, f ? f->rulenum : -1);
709 }
710
711 /*
712 * IMPORTANT: the hash function for dynamic rules must be commutative
713 * in source and destination (ip,port), because rules are bidirectional
714 * and we want to find both in the same bucket.
715 */
716 static __inline int
717 hash_packet(struct ipfw_flow_id *id)
718 {
719 u_int32_t i;
720
721 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
722 i &= (curr_dyn_buckets - 1);
723 return i;
724 }
725
726 /**
727 * unlink a dynamic rule from a chain. prev is a pointer to
728 * the previous one, q is a pointer to the rule to delete,
729 * head is a pointer to the head of the queue.
730 * Modifies q and potentially also head.
731 */
732 #define UNLINK_DYN_RULE(prev, head, q) { \
733 ipfw_dyn_rule *old_q = q; \
734 \
735 /* remove a refcount to the parent */ \
736 if (q->dyn_type == O_LIMIT) \
737 q->parent->count--; \
738 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\
739 (q->id.src_ip), (q->id.src_port), \
740 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
741 if (prev != NULL) \
742 prev->next = q = q->next; \
743 else \
744 head = q = q->next; \
745 dyn_count--; \
746 free(old_q, M_IPFW); }
747
748 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0)
749
750 /**
751 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
752 *
753 * If keep_me == NULL, rules are deleted even if not expired,
754 * otherwise only expired rules are removed.
755 *
756 * The value of the second parameter is also used to point to identify
757 * a rule we absolutely do not want to remove (e.g. because we are
758 * holding a reference to it -- this is the case with O_LIMIT_PARENT
759 * rules). The pointer is only used for comparison, so any non-null
760 * value will do.
761 */
762 static void
763 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
764 {
765 static u_int32_t last_remove = 0;
766
767 #define FORCE (keep_me == NULL)
768
769 ipfw_dyn_rule *prev, *q;
770 int i, pass = 0, max_pass = 0;
771
772 IPFW_DYN_LOCK_ASSERT();
773
774 if (ipfw_dyn_v == NULL || dyn_count == 0)
775 return;
776 /* do not expire more than once per second, it is useless */
777 if (!FORCE && last_remove == time_second)
778 return;
779 last_remove = time_second;
780
781 /*
782 * because O_LIMIT refer to parent rules, during the first pass only
783 * remove child and mark any pending LIMIT_PARENT, and remove
784 * them in a second pass.
785 */
786 next_pass:
787 for (i = 0 ; i < curr_dyn_buckets ; i++) {
788 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
789 /*
790 * Logic can become complex here, so we split tests.
791 */
792 if (q == keep_me)
793 goto next;
794 if (rule != NULL && rule != q->rule)
795 goto next; /* not the one we are looking for */
796 if (q->dyn_type == O_LIMIT_PARENT) {
797 /*
798 * handle parent in the second pass,
799 * record we need one.
800 */
801 max_pass = 1;
802 if (pass == 0)
803 goto next;
804 if (FORCE && q->count != 0 ) {
805 /* XXX should not happen! */
806 printf("ipfw: OUCH! cannot remove rule,"
807 " count %d\n", q->count);
808 }
809 } else {
810 if (!FORCE &&
811 !TIME_LEQ( q->expire, time_second ))
812 goto next;
813 }
814 if (q->dyn_type != O_LIMIT_PARENT || !q->count) {
815 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
816 continue;
817 }
818 next:
819 prev=q;
820 q=q->next;
821 }
822 }
823 if (pass++ < max_pass)
824 goto next_pass;
825 }
826
827
828 /**
829 * lookup a dynamic rule.
830 */
831 static ipfw_dyn_rule *
832 lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction,
833 struct tcphdr *tcp)
834 {
835 /*
836 * stateful ipfw extensions.
837 * Lookup into dynamic session queue
838 */
839 #define MATCH_REVERSE 0
840 #define MATCH_FORWARD 1
841 #define MATCH_NONE 2
842 #define MATCH_UNKNOWN 3
843 int i, dir = MATCH_NONE;
844 ipfw_dyn_rule *prev, *q=NULL;
845
846 IPFW_DYN_LOCK_ASSERT();
847
848 if (ipfw_dyn_v == NULL)
849 goto done; /* not found */
850 i = hash_packet( pkt );
851 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
852 if (q->dyn_type == O_LIMIT_PARENT && q->count)
853 goto next;
854 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */
855 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
856 continue;
857 }
858 if (pkt->proto == q->id.proto &&
859 q->dyn_type != O_LIMIT_PARENT) {
860 if (pkt->src_ip == q->id.src_ip &&
861 pkt->dst_ip == q->id.dst_ip &&
862 pkt->src_port == q->id.src_port &&
863 pkt->dst_port == q->id.dst_port ) {
864 dir = MATCH_FORWARD;
865 break;
866 }
867 if (pkt->src_ip == q->id.dst_ip &&
868 pkt->dst_ip == q->id.src_ip &&
869 pkt->src_port == q->id.dst_port &&
870 pkt->dst_port == q->id.src_port ) {
871 dir = MATCH_REVERSE;
872 break;
873 }
874 }
875 next:
876 prev = q;
877 q = q->next;
878 }
879 if (q == NULL)
880 goto done; /* q = NULL, not found */
881
882 if ( prev != NULL) { /* found and not in front */
883 prev->next = q->next;
884 q->next = ipfw_dyn_v[i];
885 ipfw_dyn_v[i] = q;
886 }
887 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
888 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
889
890 #define BOTH_SYN (TH_SYN | (TH_SYN << 8))
891 #define BOTH_FIN (TH_FIN | (TH_FIN << 8))
892 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
893 switch (q->state) {
894 case TH_SYN: /* opening */
895 q->expire = time_second + dyn_syn_lifetime;
896 break;
897
898 case BOTH_SYN: /* move to established */
899 case BOTH_SYN | TH_FIN : /* one side tries to close */
900 case BOTH_SYN | (TH_FIN << 8) :
901 if (tcp) {
902 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
903 u_int32_t ack = ntohl(tcp->th_ack);
904 if (dir == MATCH_FORWARD) {
905 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
906 q->ack_fwd = ack;
907 else { /* ignore out-of-sequence */
908 break;
909 }
910 } else {
911 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
912 q->ack_rev = ack;
913 else { /* ignore out-of-sequence */
914 break;
915 }
916 }
917 }
918 q->expire = time_second + dyn_ack_lifetime;
919 break;
920
921 case BOTH_SYN | BOTH_FIN: /* both sides closed */
922 if (dyn_fin_lifetime >= dyn_keepalive_period)
923 dyn_fin_lifetime = dyn_keepalive_period - 1;
924 q->expire = time_second + dyn_fin_lifetime;
925 break;
926
927 default:
928 #if 0
929 /*
930 * reset or some invalid combination, but can also
931 * occur if we use keep-state the wrong way.
932 */
933 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
934 printf("invalid state: 0x%x\n", q->state);
935 #endif
936 if (dyn_rst_lifetime >= dyn_keepalive_period)
937 dyn_rst_lifetime = dyn_keepalive_period - 1;
938 q->expire = time_second + dyn_rst_lifetime;
939 break;
940 }
941 } else if (pkt->proto == IPPROTO_UDP) {
942 q->expire = time_second + dyn_udp_lifetime;
943 } else {
944 /* other protocols */
945 q->expire = time_second + dyn_short_lifetime;
946 }
947 done:
948 if (match_direction)
949 *match_direction = dir;
950 return q;
951 }
952
953 static ipfw_dyn_rule *
954 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
955 struct tcphdr *tcp)
956 {
957 ipfw_dyn_rule *q;
958
959 IPFW_DYN_LOCK();
960 q = lookup_dyn_rule_locked(pkt, match_direction, tcp);
961 if (q == NULL)
962 IPFW_DYN_UNLOCK();
963 /* NB: return table locked when q is not NULL */
964 return q;
965 }
966
967 static void
968 realloc_dynamic_table(void)
969 {
970 IPFW_DYN_LOCK_ASSERT();
971
972 /*
973 * Try reallocation, make sure we have a power of 2 and do
974 * not allow more than 64k entries. In case of overflow,
975 * default to 1024.
976 */
977
978 if (dyn_buckets > 65536)
979 dyn_buckets = 1024;
980 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
981 dyn_buckets = curr_dyn_buckets; /* reset */
982 return;
983 }
984 curr_dyn_buckets = dyn_buckets;
985 if (ipfw_dyn_v != NULL)
986 free(ipfw_dyn_v, M_IPFW);
987 for (;;) {
988 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
989 M_IPFW, M_NOWAIT | M_ZERO);
990 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
991 break;
992 curr_dyn_buckets /= 2;
993 }
994 }
995
996 /**
997 * Install state of type 'type' for a dynamic session.
998 * The hash table contains two type of rules:
999 * - regular rules (O_KEEP_STATE)
1000 * - rules for sessions with limited number of sess per user
1001 * (O_LIMIT). When they are created, the parent is
1002 * increased by 1, and decreased on delete. In this case,
1003 * the third parameter is the parent rule and not the chain.
1004 * - "parent" rules for the above (O_LIMIT_PARENT).
1005 */
1006 static ipfw_dyn_rule *
1007 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
1008 {
1009 ipfw_dyn_rule *r;
1010 int i;
1011
1012 IPFW_DYN_LOCK_ASSERT();
1013
1014 if (ipfw_dyn_v == NULL ||
1015 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
1016 realloc_dynamic_table();
1017 if (ipfw_dyn_v == NULL)
1018 return NULL; /* failed ! */
1019 }
1020 i = hash_packet(id);
1021
1022 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO);
1023 if (r == NULL) {
1024 printf ("ipfw: sorry cannot allocate state\n");
1025 return NULL;
1026 }
1027
1028 /* increase refcount on parent, and set pointer */
1029 if (dyn_type == O_LIMIT) {
1030 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
1031 if ( parent->dyn_type != O_LIMIT_PARENT)
1032 panic("invalid parent");
1033 parent->count++;
1034 r->parent = parent;
1035 rule = parent->rule;
1036 }
1037
1038 r->id = *id;
1039 r->expire = time_second + dyn_syn_lifetime;
1040 r->rule = rule;
1041 r->dyn_type = dyn_type;
1042 r->pcnt = r->bcnt = 0;
1043 r->count = 0;
1044
1045 r->bucket = i;
1046 r->next = ipfw_dyn_v[i];
1047 ipfw_dyn_v[i] = r;
1048 dyn_count++;
1049 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
1050 dyn_type,
1051 (r->id.src_ip), (r->id.src_port),
1052 (r->id.dst_ip), (r->id.dst_port),
1053 dyn_count ); )
1054 return r;
1055 }
1056
1057 /**
1058 * lookup dynamic parent rule using pkt and rule as search keys.
1059 * If the lookup fails, then install one.
1060 */
1061 static ipfw_dyn_rule *
1062 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
1063 {
1064 ipfw_dyn_rule *q;
1065 int i;
1066
1067 IPFW_DYN_LOCK_ASSERT();
1068
1069 if (ipfw_dyn_v) {
1070 i = hash_packet( pkt );
1071 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
1072 if (q->dyn_type == O_LIMIT_PARENT &&
1073 rule== q->rule &&
1074 pkt->proto == q->id.proto &&
1075 pkt->src_ip == q->id.src_ip &&
1076 pkt->dst_ip == q->id.dst_ip &&
1077 pkt->src_port == q->id.src_port &&
1078 pkt->dst_port == q->id.dst_port) {
1079 q->expire = time_second + dyn_short_lifetime;
1080 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);)
1081 return q;
1082 }
1083 }
1084 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
1085 }
1086
1087 /**
1088 * Install dynamic state for rule type cmd->o.opcode
1089 *
1090 * Returns 1 (failure) if state is not installed because of errors or because
1091 * session limitations are enforced.
1092 */
1093 static int
1094 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
1095 struct ip_fw_args *args)
1096 {
1097 static int last_log;
1098
1099 ipfw_dyn_rule *q;
1100
1101 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n",
1102 cmd->o.opcode,
1103 (args->f_id.src_ip), (args->f_id.src_port),
1104 (args->f_id.dst_ip), (args->f_id.dst_port) );)
1105
1106 IPFW_DYN_LOCK();
1107
1108 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1109
1110 if (q != NULL) { /* should never occur */
1111 if (last_log != time_second) {
1112 last_log = time_second;
1113 printf("ipfw: install_state: entry already present, done\n");
1114 }
1115 IPFW_DYN_UNLOCK();
1116 return 0;
1117 }
1118
1119 if (dyn_count >= dyn_max)
1120 /*
1121 * Run out of slots, try to remove any expired rule.
1122 */
1123 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1124
1125 if (dyn_count >= dyn_max) {
1126 if (last_log != time_second) {
1127 last_log = time_second;
1128 printf("ipfw: install_state: Too many dynamic rules\n");
1129 }
1130 IPFW_DYN_UNLOCK();
1131 return 1; /* cannot install, notify caller */
1132 }
1133
1134 switch (cmd->o.opcode) {
1135 case O_KEEP_STATE: /* bidir rule */
1136 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1137 break;
1138
1139 case O_LIMIT: /* limit number of sessions */
1140 {
1141 u_int16_t limit_mask = cmd->limit_mask;
1142 struct ipfw_flow_id id;
1143 ipfw_dyn_rule *parent;
1144
1145 DEB(printf("ipfw: installing dyn-limit rule %d\n",
1146 cmd->conn_limit);)
1147
1148 id.dst_ip = id.src_ip = 0;
1149 id.dst_port = id.src_port = 0;
1150 id.proto = args->f_id.proto;
1151
1152 if (limit_mask & DYN_SRC_ADDR)
1153 id.src_ip = args->f_id.src_ip;
1154 if (limit_mask & DYN_DST_ADDR)
1155 id.dst_ip = args->f_id.dst_ip;
1156 if (limit_mask & DYN_SRC_PORT)
1157 id.src_port = args->f_id.src_port;
1158 if (limit_mask & DYN_DST_PORT)
1159 id.dst_port = args->f_id.dst_port;
1160 parent = lookup_dyn_parent(&id, rule);
1161 if (parent == NULL) {
1162 printf("ipfw: add parent failed\n");
1163 return 1;
1164 }
1165 if (parent->count >= cmd->conn_limit) {
1166 /*
1167 * See if we can remove some expired rule.
1168 */
1169 remove_dyn_rule(rule, parent);
1170 if (parent->count >= cmd->conn_limit) {
1171 if (fw_verbose && last_log != time_second) {
1172 last_log = time_second;
1173 log(LOG_SECURITY | LOG_DEBUG,
1174 "drop session, too many entries\n");
1175 }
1176 IPFW_DYN_UNLOCK();
1177 return 1;
1178 }
1179 }
1180 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent);
1181 }
1182 break;
1183 default:
1184 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode);
1185 IPFW_DYN_UNLOCK();
1186 return 1;
1187 }
1188 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */
1189 IPFW_DYN_UNLOCK();
1190 return 0;
1191 }
1192
1193 /*
1194 * Transmit a TCP packet, containing either a RST or a keepalive.
1195 * When flags & TH_RST, we are sending a RST packet, because of a
1196 * "reset" action matched the packet.
1197 * Otherwise we are sending a keepalive, and flags & TH_
1198 */
1199 static void
1200 send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags)
1201 {
1202 struct mbuf *m;
1203 struct ip *ip;
1204 struct tcphdr *tcp;
1205
1206 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1207 if (m == 0)
1208 return;
1209 m->m_pkthdr.rcvif = (struct ifnet *)0;
1210 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1211 m->m_data += max_linkhdr;
1212
1213 ip = mtod(m, struct ip *);
1214 bzero(ip, m->m_len);
1215 tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1216 ip->ip_p = IPPROTO_TCP;
1217 tcp->th_off = 5;
1218 /*
1219 * Assume we are sending a RST (or a keepalive in the reverse
1220 * direction), swap src and destination addresses and ports.
1221 */
1222 ip->ip_src.s_addr = htonl(id->dst_ip);
1223 ip->ip_dst.s_addr = htonl(id->src_ip);
1224 tcp->th_sport = htons(id->dst_port);
1225 tcp->th_dport = htons(id->src_port);
1226 if (flags & TH_RST) { /* we are sending a RST */
1227 if (flags & TH_ACK) {
1228 tcp->th_seq = htonl(ack);
1229 tcp->th_ack = htonl(0);
1230 tcp->th_flags = TH_RST;
1231 } else {
1232 if (flags & TH_SYN)
1233 seq++;
1234 tcp->th_seq = htonl(0);
1235 tcp->th_ack = htonl(seq);
1236 tcp->th_flags = TH_RST | TH_ACK;
1237 }
1238 } else {
1239 /*
1240 * We are sending a keepalive. flags & TH_SYN determines
1241 * the direction, forward if set, reverse if clear.
1242 * NOTE: seq and ack are always assumed to be correct
1243 * as set by the caller. This may be confusing...
1244 */
1245 if (flags & TH_SYN) {
1246 /*
1247 * we have to rewrite the correct addresses!
1248 */
1249 ip->ip_dst.s_addr = htonl(id->dst_ip);
1250 ip->ip_src.s_addr = htonl(id->src_ip);
1251 tcp->th_dport = htons(id->dst_port);
1252 tcp->th_sport = htons(id->src_port);
1253 }
1254 tcp->th_seq = htonl(seq);
1255 tcp->th_ack = htonl(ack);
1256 tcp->th_flags = TH_ACK;
1257 }
1258 /*
1259 * set ip_len to the payload size so we can compute
1260 * the tcp checksum on the pseudoheader
1261 * XXX check this, could save a couple of words ?
1262 */
1263 ip->ip_len = htons(sizeof(struct tcphdr));
1264 tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1265 /*
1266 * now fill fields left out earlier
1267 */
1268 ip->ip_ttl = ip_defttl;
1269 ip->ip_len = m->m_pkthdr.len;
1270 m->m_flags |= M_SKIP_FIREWALL;
1271 ip_output(m, NULL, NULL, 0, NULL, NULL);
1272 }
1273
1274 /*
1275 * sends a reject message, consuming the mbuf passed as an argument.
1276 */
1277 static void
1278 send_reject(struct ip_fw_args *args, int code, int offset, int ip_len)
1279 {
1280
1281 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1282 /* We need the IP header in host order for icmp_error(). */
1283 if (args->eh != NULL) {
1284 struct ip *ip = mtod(args->m, struct ip *);
1285 ip->ip_len = ntohs(ip->ip_len);
1286 ip->ip_off = ntohs(ip->ip_off);
1287 }
1288 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1289 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) {
1290 struct tcphdr *const tcp =
1291 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1292 if ( (tcp->th_flags & TH_RST) == 0)
1293 send_pkt(&(args->f_id), ntohl(tcp->th_seq),
1294 ntohl(tcp->th_ack),
1295 tcp->th_flags | TH_RST);
1296 m_freem(args->m);
1297 } else
1298 m_freem(args->m);
1299 args->m = NULL;
1300 }
1301
1302 /**
1303 *
1304 * Given an ip_fw *, lookup_next_rule will return a pointer
1305 * to the next rule, which can be either the jump
1306 * target (for skipto instructions) or the next one in the list (in
1307 * all other cases including a missing jump target).
1308 * The result is also written in the "next_rule" field of the rule.
1309 * Backward jumps are not allowed, so start looking from the next
1310 * rule...
1311 *
1312 * This never returns NULL -- in case we do not have an exact match,
1313 * the next rule is returned. When the ruleset is changed,
1314 * pointers are flushed so we are always correct.
1315 */
1316
1317 static struct ip_fw *
1318 lookup_next_rule(struct ip_fw *me)
1319 {
1320 struct ip_fw *rule = NULL;
1321 ipfw_insn *cmd;
1322
1323 /* look for action, in case it is a skipto */
1324 cmd = ACTION_PTR(me);
1325 if (cmd->opcode == O_LOG)
1326 cmd += F_LEN(cmd);
1327 if ( cmd->opcode == O_SKIPTO )
1328 for (rule = me->next; rule ; rule = rule->next)
1329 if (rule->rulenum >= cmd->arg1)
1330 break;
1331 if (rule == NULL) /* failure or not a skipto */
1332 rule = me->next;
1333 me->next_rule = rule;
1334 return rule;
1335 }
1336
1337 static void
1338 init_tables(void)
1339 {
1340 int i;
1341
1342 for (i = 0; i < IPFW_TABLES_MAX; i++) {
1343 rn_inithead((void **)&ipfw_tables[i].rnh, 32);
1344 ipfw_tables[i].modified = 1;
1345 }
1346 }
1347
1348 static int
1349 add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value)
1350 {
1351 struct radix_node_head *rnh;
1352 struct table_entry *ent;
1353
1354 if (tbl >= IPFW_TABLES_MAX)
1355 return (EINVAL);
1356 rnh = ipfw_tables[tbl].rnh;
1357 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO);
1358 if (ent == NULL)
1359 return (ENOMEM);
1360 ent->value = value;
1361 ent->addr.sin_len = ent->mask.sin_len = 8;
1362 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1363 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr;
1364 RADIX_NODE_HEAD_LOCK(rnh);
1365 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) ==
1366 NULL) {
1367 RADIX_NODE_HEAD_UNLOCK(rnh);
1368 free(ent, M_IPFW_TBL);
1369 return (EEXIST);
1370 }
1371 ipfw_tables[tbl].modified = 1;
1372 RADIX_NODE_HEAD_UNLOCK(rnh);
1373 return (0);
1374 }
1375
1376 static int
1377 del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen)
1378 {
1379 struct radix_node_head *rnh;
1380 struct table_entry *ent;
1381 struct sockaddr_in sa, mask;
1382
1383 if (tbl >= IPFW_TABLES_MAX)
1384 return (EINVAL);
1385 rnh = ipfw_tables[tbl].rnh;
1386 sa.sin_len = mask.sin_len = 8;
1387 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1388 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr;
1389 RADIX_NODE_HEAD_LOCK(rnh);
1390 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh);
1391 if (ent == NULL) {
1392 RADIX_NODE_HEAD_UNLOCK(rnh);
1393 return (ESRCH);
1394 }
1395 ipfw_tables[tbl].modified = 1;
1396 RADIX_NODE_HEAD_UNLOCK(rnh);
1397 free(ent, M_IPFW_TBL);
1398 return (0);
1399 }
1400
1401 static int
1402 flush_table_entry(struct radix_node *rn, void *arg)
1403 {
1404 struct radix_node_head * const rnh = arg;
1405 struct table_entry *ent;
1406
1407 ent = (struct table_entry *)
1408 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
1409 if (ent != NULL)
1410 free(ent, M_IPFW_TBL);
1411 return (0);
1412 }
1413
1414 static int
1415 flush_table(u_int16_t tbl)
1416 {
1417 struct radix_node_head *rnh;
1418
1419 if (tbl >= IPFW_TABLES_MAX)
1420 return (EINVAL);
1421 rnh = ipfw_tables[tbl].rnh;
1422 RADIX_NODE_HEAD_LOCK(rnh);
1423 rnh->rnh_walktree(rnh, flush_table_entry, rnh);
1424 ipfw_tables[tbl].modified = 1;
1425 RADIX_NODE_HEAD_UNLOCK(rnh);
1426 return (0);
1427 }
1428
1429 static void
1430 flush_tables(void)
1431 {
1432 u_int16_t tbl;
1433
1434 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++)
1435 flush_table(tbl);
1436 }
1437
1438 static int
1439 lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val)
1440 {
1441 struct radix_node_head *rnh;
1442 struct table_entry *ent;
1443 struct sockaddr_in sa;
1444 static in_addr_t last_addr;
1445 static int last_tbl;
1446 static int last_match;
1447 static u_int32_t last_value;
1448
1449 if (tbl >= IPFW_TABLES_MAX)
1450 return (0);
1451 if (tbl == last_tbl && addr == last_addr &&
1452 !ipfw_tables[tbl].modified) {
1453 if (last_match)
1454 *val = last_value;
1455 return (last_match);
1456 }
1457 rnh = ipfw_tables[tbl].rnh;
1458 sa.sin_len = 8;
1459 sa.sin_addr.s_addr = addr;
1460 RADIX_NODE_HEAD_LOCK(rnh);
1461 ipfw_tables[tbl].modified = 0;
1462 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh));
1463 RADIX_NODE_HEAD_UNLOCK(rnh);
1464 last_addr = addr;
1465 last_tbl = tbl;
1466 if (ent != NULL) {
1467 last_value = *val = ent->value;
1468 last_match = 1;
1469 return (1);
1470 }
1471 last_match = 0;
1472 return (0);
1473 }
1474
1475 static int
1476 count_table_entry(struct radix_node *rn, void *arg)
1477 {
1478 u_int32_t * const cnt = arg;
1479
1480 (*cnt)++;
1481 return (0);
1482 }
1483
1484 static int
1485 count_table(u_int32_t tbl, u_int32_t *cnt)
1486 {
1487 struct radix_node_head *rnh;
1488
1489 if (tbl >= IPFW_TABLES_MAX)
1490 return (EINVAL);
1491 rnh = ipfw_tables[tbl].rnh;
1492 *cnt = 0;
1493 RADIX_NODE_HEAD_LOCK(rnh);
1494 rnh->rnh_walktree(rnh, count_table_entry, cnt);
1495 RADIX_NODE_HEAD_UNLOCK(rnh);
1496 return (0);
1497 }
1498
1499 static int
1500 dump_table_entry(struct radix_node *rn, void *arg)
1501 {
1502 struct table_entry * const n = (struct table_entry *)rn;
1503 ipfw_table * const tbl = arg;
1504 ipfw_table_entry *ent;
1505
1506 if (tbl->cnt == tbl->size)
1507 return (1);
1508 ent = &tbl->ent[tbl->cnt];
1509 ent->tbl = tbl->tbl;
1510 if (in_nullhost(n->mask.sin_addr))
1511 ent->masklen = 0;
1512 else
1513 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr));
1514 ent->addr = n->addr.sin_addr.s_addr;
1515 ent->value = n->value;
1516 tbl->cnt++;
1517 return (0);
1518 }
1519
1520 static int
1521 dump_table(ipfw_table *tbl)
1522 {
1523 struct radix_node_head *rnh;
1524
1525 if (tbl->tbl >= IPFW_TABLES_MAX)
1526 return (EINVAL);
1527 rnh = ipfw_tables[tbl->tbl].rnh;
1528 tbl->cnt = 0;
1529 RADIX_NODE_HEAD_LOCK(rnh);
1530 rnh->rnh_walktree(rnh, dump_table_entry, tbl);
1531 RADIX_NODE_HEAD_UNLOCK(rnh);
1532 return (0);
1533 }
1534
1535 static void
1536 fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp)
1537 {
1538 struct ucred *cr;
1539
1540 if (inp->inp_socket != NULL) {
1541 cr = inp->inp_socket->so_cred;
1542 ugp->fw_prid = jailed(cr) ?
1543 cr->cr_prison->pr_id : -1;
1544 ugp->fw_uid = cr->cr_uid;
1545 ugp->fw_ngroups = cr->cr_ngroups;
1546 bcopy(cr->cr_groups, ugp->fw_groups,
1547 sizeof(ugp->fw_groups));
1548 }
1549 }
1550
1551 static int
1552 check_uidgid(ipfw_insn_u32 *insn,
1553 int proto, struct ifnet *oif,
1554 struct in_addr dst_ip, u_int16_t dst_port,
1555 struct in_addr src_ip, u_int16_t src_port,
1556 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp)
1557 {
1558 struct inpcbinfo *pi;
1559 int wildcard;
1560 struct inpcb *pcb;
1561 int match;
1562 gid_t *gp;
1563
1564 /*
1565 * Check to see if the UDP or TCP stack supplied us with
1566 * the PCB. If so, rather then holding a lock and looking
1567 * up the PCB, we can use the one that was supplied.
1568 */
1569 if (inp && *lookup == 0) {
1570 INP_LOCK_ASSERT(inp);
1571 if (inp->inp_socket != NULL) {
1572 fill_ugid_cache(inp, ugp);
1573 *lookup = 1;
1574 }
1575 }
1576 /*
1577 * If we have already been here and the packet has no
1578 * PCB entry associated with it, then we can safely
1579 * assume that this is a no match.
1580 */
1581 if (*lookup == -1)
1582 return (0);
1583 if (proto == IPPROTO_TCP) {
1584 wildcard = 0;
1585 pi = &tcbinfo;
1586 } else if (proto == IPPROTO_UDP) {
1587 wildcard = 1;
1588 pi = &udbinfo;
1589 } else
1590 return 0;
1591 match = 0;
1592 if (*lookup == 0) {
1593 INP_INFO_RLOCK(pi);
1594 pcb = (oif) ?
1595 in_pcblookup_hash(pi,
1596 dst_ip, htons(dst_port),
1597 src_ip, htons(src_port),
1598 wildcard, oif) :
1599 in_pcblookup_hash(pi,
1600 src_ip, htons(src_port),
1601 dst_ip, htons(dst_port),
1602 wildcard, NULL);
1603 if (pcb != NULL) {
1604 INP_LOCK(pcb);
1605 if (pcb->inp_socket != NULL) {
1606 fill_ugid_cache(pcb, ugp);
1607 *lookup = 1;
1608 }
1609 INP_UNLOCK(pcb);
1610 }
1611 INP_INFO_RUNLOCK(pi);
1612 if (*lookup == 0) {
1613 /*
1614 * If the lookup did not yield any results, there
1615 * is no sense in coming back and trying again. So
1616 * we can set lookup to -1 and ensure that we wont
1617 * bother the pcb system again.
1618 */
1619 *lookup = -1;
1620 return (0);
1621 }
1622 }
1623 if (insn->o.opcode == O_UID)
1624 match = (ugp->fw_uid == (uid_t)insn->d[0]);
1625 else if (insn->o.opcode == O_GID) {
1626 for (gp = ugp->fw_groups;
1627 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++)
1628 if (*gp == (gid_t)insn->d[0]) {
1629 match = 1;
1630 break;
1631 }
1632 } else if (insn->o.opcode == O_JAIL)
1633 match = (ugp->fw_prid == (int)insn->d[0]);
1634 return match;
1635 }
1636
1637 /*
1638 * The main check routine for the firewall.
1639 *
1640 * All arguments are in args so we can modify them and return them
1641 * back to the caller.
1642 *
1643 * Parameters:
1644 *
1645 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
1646 * Starts with the IP header.
1647 * args->eh (in) Mac header if present, or NULL for layer3 packet.
1648 * args->oif Outgoing interface, or NULL if packet is incoming.
1649 * The incoming interface is in the mbuf. (in)
1650 * args->divert_rule (in/out)
1651 * Skip up to the first rule past this rule number;
1652 * upon return, non-zero port number for divert or tee.
1653 *
1654 * args->rule Pointer to the last matching rule (in/out)
1655 * args->next_hop Socket we are forwarding to (out).
1656 * args->f_id Addresses grabbed from the packet (out)
1657 *
1658 * Return value:
1659 *
1660 * IP_FW_PORT_DENY_FLAG the packet must be dropped.
1661 * 0 The packet is to be accepted and routed normally OR
1662 * the packet was denied/rejected and has been dropped;
1663 * in the latter case, *m is equal to NULL upon return.
1664 * port Divert the packet to port, with these caveats:
1665 *
1666 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead
1667 * of diverting it (ie, 'ipfw tee').
1668 *
1669 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower
1670 * 16 bits as a dummynet pipe number instead of diverting
1671 */
1672
1673 int
1674 ipfw_chk(struct ip_fw_args *args)
1675 {
1676 /*
1677 * Local variables hold state during the processing of a packet.
1678 *
1679 * IMPORTANT NOTE: to speed up the processing of rules, there
1680 * are some assumption on the values of the variables, which
1681 * are documented here. Should you change them, please check
1682 * the implementation of the various instructions to make sure
1683 * that they still work.
1684 *
1685 * args->eh The MAC header. It is non-null for a layer2
1686 * packet, it is NULL for a layer-3 packet.
1687 *
1688 * m | args->m Pointer to the mbuf, as received from the caller.
1689 * It may change if ipfw_chk() does an m_pullup, or if it
1690 * consumes the packet because it calls send_reject().
1691 * XXX This has to change, so that ipfw_chk() never modifies
1692 * or consumes the buffer.
1693 * ip is simply an alias of the value of m, and it is kept
1694 * in sync with it (the packet is supposed to start with
1695 * the ip header).
1696 */
1697 struct mbuf *m = args->m;
1698 struct ip *ip = mtod(m, struct ip *);
1699
1700 /*
1701 * For rules which contain uid/gid or jail constraints, cache
1702 * a copy of the users credentials after the pcb lookup has been
1703 * executed. This will speed up the processing of rules with
1704 * these types of constraints, as well as decrease contention
1705 * on pcb related locks.
1706 */
1707 struct ip_fw_ugid fw_ugid_cache;
1708 int ugid_lookup = 0;
1709
1710 /*
1711 * oif | args->oif If NULL, ipfw_chk has been called on the
1712 * inbound path (ether_input, bdg_forward, ip_input).
1713 * If non-NULL, ipfw_chk has been called on the outbound path
1714 * (ether_output, ip_output).
1715 */
1716 struct ifnet *oif = args->oif;
1717
1718 struct ip_fw *f = NULL; /* matching rule */
1719 int retval = 0;
1720
1721 /*
1722 * hlen The length of the IPv4 header.
1723 * hlen >0 means we have an IPv4 packet.
1724 */
1725 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
1726
1727 /*
1728 * offset The offset of a fragment. offset != 0 means that
1729 * we have a fragment at this offset of an IPv4 packet.
1730 * offset == 0 means that (if this is an IPv4 packet)
1731 * this is the first or only fragment.
1732 */
1733 u_short offset = 0;
1734
1735 /*
1736 * Local copies of addresses. They are only valid if we have
1737 * an IP packet.
1738 *
1739 * proto The protocol. Set to 0 for non-ip packets,
1740 * or to the protocol read from the packet otherwise.
1741 * proto != 0 means that we have an IPv4 packet.
1742 *
1743 * src_port, dst_port port numbers, in HOST format. Only
1744 * valid for TCP and UDP packets.
1745 *
1746 * src_ip, dst_ip ip addresses, in NETWORK format.
1747 * Only valid for IPv4 packets.
1748 */
1749 u_int8_t proto;
1750 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */
1751 struct in_addr src_ip, dst_ip; /* NOTE: network format */
1752 u_int16_t ip_len=0;
1753 int pktlen;
1754 int dyn_dir = MATCH_UNKNOWN;
1755 ipfw_dyn_rule *q = NULL;
1756 struct ip_fw_chain *chain = &layer3_chain;
1757 struct m_tag *mtag;
1758
1759 if (m->m_flags & M_SKIP_FIREWALL)
1760 return 0; /* accept */
1761 /*
1762 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1763 * MATCH_NONE when checked and not matched (q = NULL),
1764 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1765 */
1766
1767 pktlen = m->m_pkthdr.len;
1768 if (args->eh == NULL || /* layer 3 packet */
1769 ( m->m_pkthdr.len >= sizeof(struct ip) &&
1770 ntohs(args->eh->ether_type) == ETHERTYPE_IP))
1771 hlen = ip->ip_hl << 2;
1772
1773 /*
1774 * Collect parameters into local variables for faster matching.
1775 */
1776 if (hlen == 0) { /* do not grab addresses for non-ip pkts */
1777 proto = args->f_id.proto = 0; /* mark f_id invalid */
1778 goto after_ip_checks;
1779 }
1780
1781 proto = args->f_id.proto = ip->ip_p;
1782 src_ip = ip->ip_src;
1783 dst_ip = ip->ip_dst;
1784 if (args->eh != NULL) { /* layer 2 packets are as on the wire */
1785 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1786 ip_len = ntohs(ip->ip_len);
1787 } else {
1788 offset = ip->ip_off & IP_OFFMASK;
1789 ip_len = ip->ip_len;
1790 }
1791 pktlen = ip_len < pktlen ? ip_len : pktlen;
1792
1793 #define PULLUP_TO(len) \
1794 do { \
1795 if ((m)->m_len < (len)) { \
1796 args->m = m = m_pullup(m, (len)); \
1797 if (m == 0) \
1798 goto pullup_failed; \
1799 ip = mtod(m, struct ip *); \
1800 } \
1801 } while (0)
1802
1803 if (offset == 0) {
1804 switch (proto) {
1805 case IPPROTO_TCP:
1806 {
1807 struct tcphdr *tcp;
1808
1809 PULLUP_TO(hlen + sizeof(struct tcphdr));
1810 tcp = L3HDR(struct tcphdr, ip);
1811 dst_port = tcp->th_dport;
1812 src_port = tcp->th_sport;
1813 args->f_id.flags = tcp->th_flags;
1814 }
1815 break;
1816
1817 case IPPROTO_UDP:
1818 {
1819 struct udphdr *udp;
1820
1821 PULLUP_TO(hlen + sizeof(struct udphdr));
1822 udp = L3HDR(struct udphdr, ip);
1823 dst_port = udp->uh_dport;
1824 src_port = udp->uh_sport;
1825 }
1826 break;
1827
1828 case IPPROTO_ICMP:
1829 PULLUP_TO(hlen + 4); /* type, code and checksum. */
1830 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type;
1831 break;
1832
1833 default:
1834 break;
1835 }
1836 #undef PULLUP_TO
1837 }
1838
1839 args->f_id.src_ip = ntohl(src_ip.s_addr);
1840 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1841 args->f_id.src_port = src_port = ntohs(src_port);
1842 args->f_id.dst_port = dst_port = ntohs(dst_port);
1843
1844 after_ip_checks:
1845 IPFW_LOCK(chain); /* XXX expensive? can we run lock free? */
1846 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
1847 if (args->rule) {
1848 /*
1849 * Packet has already been tagged. Look for the next rule
1850 * to restart processing.
1851 *
1852 * If fw_one_pass != 0 then just accept it.
1853 * XXX should not happen here, but optimized out in
1854 * the caller.
1855 */
1856 if (fw_one_pass) {
1857 IPFW_UNLOCK(chain); /* XXX optimize */
1858 return 0;
1859 }
1860
1861 f = args->rule->next_rule;
1862 if (f == NULL)
1863 f = lookup_next_rule(args->rule);
1864 } else {
1865 /*
1866 * Find the starting rule. It can be either the first
1867 * one, or the one after divert_rule if asked so.
1868 */
1869 int skipto = mtag ? divert_cookie(mtag) : 0;
1870
1871 f = chain->rules;
1872 if (args->eh == NULL && skipto != 0) {
1873 if (skipto >= IPFW_DEFAULT_RULE) {
1874 IPFW_UNLOCK(chain);
1875 return(IP_FW_PORT_DENY_FLAG); /* invalid */
1876 }
1877 while (f && f->rulenum <= skipto)
1878 f = f->next;
1879 if (f == NULL) { /* drop packet */
1880 IPFW_UNLOCK(chain);
1881 return(IP_FW_PORT_DENY_FLAG);
1882 }
1883 }
1884 }
1885 /* reset divert rule to avoid confusion later */
1886 if (mtag)
1887 m_tag_delete(m, mtag);
1888
1889 /*
1890 * Now scan the rules, and parse microinstructions for each rule.
1891 */
1892 for (; f; f = f->next) {
1893 int l, cmdlen;
1894 ipfw_insn *cmd;
1895 int skip_or; /* skip rest of OR block */
1896
1897 again:
1898 if (set_disable & (1 << f->set) )
1899 continue;
1900
1901 skip_or = 0;
1902 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1903 l -= cmdlen, cmd += cmdlen) {
1904 int match;
1905
1906 /*
1907 * check_body is a jump target used when we find a
1908 * CHECK_STATE, and need to jump to the body of
1909 * the target rule.
1910 */
1911
1912 check_body:
1913 cmdlen = F_LEN(cmd);
1914 /*
1915 * An OR block (insn_1 || .. || insn_n) has the
1916 * F_OR bit set in all but the last instruction.
1917 * The first match will set "skip_or", and cause
1918 * the following instructions to be skipped until
1919 * past the one with the F_OR bit clear.
1920 */
1921 if (skip_or) { /* skip this instruction */
1922 if ((cmd->len & F_OR) == 0)
1923 skip_or = 0; /* next one is good */
1924 continue;
1925 }
1926 match = 0; /* set to 1 if we succeed */
1927
1928 switch (cmd->opcode) {
1929 /*
1930 * The first set of opcodes compares the packet's
1931 * fields with some pattern, setting 'match' if a
1932 * match is found. At the end of the loop there is
1933 * logic to deal with F_NOT and F_OR flags associated
1934 * with the opcode.
1935 */
1936 case O_NOP:
1937 match = 1;
1938 break;
1939
1940 case O_FORWARD_MAC:
1941 printf("ipfw: opcode %d unimplemented\n",
1942 cmd->opcode);
1943 break;
1944
1945 case O_GID:
1946 case O_UID:
1947 case O_JAIL:
1948 /*
1949 * We only check offset == 0 && proto != 0,
1950 * as this ensures that we have an IPv4
1951 * packet with the ports info.
1952 */
1953 if (offset!=0)
1954 break;
1955 if (proto == IPPROTO_TCP ||
1956 proto == IPPROTO_UDP)
1957 match = check_uidgid(
1958 (ipfw_insn_u32 *)cmd,
1959 proto, oif,
1960 dst_ip, dst_port,
1961 src_ip, src_port, &fw_ugid_cache,
1962 &ugid_lookup, args->inp);
1963 break;
1964
1965 case O_RECV:
1966 match = iface_match(m->m_pkthdr.rcvif,
1967 (ipfw_insn_if *)cmd);
1968 break;
1969
1970 case O_XMIT:
1971 match = iface_match(oif, (ipfw_insn_if *)cmd);
1972 break;
1973
1974 case O_VIA:
1975 match = iface_match(oif ? oif :
1976 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
1977 break;
1978
1979 case O_MACADDR2:
1980 if (args->eh != NULL) { /* have MAC header */
1981 u_int32_t *want = (u_int32_t *)
1982 ((ipfw_insn_mac *)cmd)->addr;
1983 u_int32_t *mask = (u_int32_t *)
1984 ((ipfw_insn_mac *)cmd)->mask;
1985 u_int32_t *hdr = (u_int32_t *)args->eh;
1986
1987 match =
1988 ( want[0] == (hdr[0] & mask[0]) &&
1989 want[1] == (hdr[1] & mask[1]) &&
1990 want[2] == (hdr[2] & mask[2]) );
1991 }
1992 break;
1993
1994 case O_MAC_TYPE:
1995 if (args->eh != NULL) {
1996 u_int16_t t =
1997 ntohs(args->eh->ether_type);
1998 u_int16_t *p =
1999 ((ipfw_insn_u16 *)cmd)->ports;
2000 int i;
2001
2002 for (i = cmdlen - 1; !match && i>0;
2003 i--, p += 2)
2004 match = (t>=p[0] && t<=p[1]);
2005 }
2006 break;
2007
2008 case O_FRAG:
2009 match = (hlen > 0 && offset != 0);
2010 break;
2011
2012 case O_IN: /* "out" is "not in" */
2013 match = (oif == NULL);
2014 break;
2015
2016 case O_LAYER2:
2017 match = (args->eh != NULL);
2018 break;
2019
2020 case O_PROTO:
2021 /*
2022 * We do not allow an arg of 0 so the
2023 * check of "proto" only suffices.
2024 */
2025 match = (proto == cmd->arg1);
2026 break;
2027
2028 case O_IP_SRC:
2029 match = (hlen > 0 &&
2030 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
2031 src_ip.s_addr);
2032 break;
2033
2034 case O_IP_SRC_LOOKUP:
2035 case O_IP_DST_LOOKUP:
2036 if (hlen > 0) {
2037 uint32_t a =
2038 (cmd->opcode == O_IP_DST_LOOKUP) ?
2039 dst_ip.s_addr : src_ip.s_addr;
2040 uint32_t v;
2041
2042 match = lookup_table(cmd->arg1, a, &v);
2043 if (!match)
2044 break;
2045 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2046 match =
2047 ((ipfw_insn_u32 *)cmd)->d[0] == v;
2048 }
2049 break;
2050
2051 case O_IP_SRC_MASK:
2052 case O_IP_DST_MASK:
2053 if (hlen > 0) {
2054 uint32_t a =
2055 (cmd->opcode == O_IP_DST_MASK) ?
2056 dst_ip.s_addr : src_ip.s_addr;
2057 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2058 int i = cmdlen-1;
2059
2060 for (; !match && i>0; i-= 2, p+= 2)
2061 match = (p[0] == (a & p[1]));
2062 }
2063 break;
2064
2065 case O_IP_SRC_ME:
2066 if (hlen > 0) {
2067 struct ifnet *tif;
2068
2069 INADDR_TO_IFP(src_ip, tif);
2070 match = (tif != NULL);
2071 }
2072 break;
2073
2074 case O_IP_DST_SET:
2075 case O_IP_SRC_SET:
2076 if (hlen > 0) {
2077 u_int32_t *d = (u_int32_t *)(cmd+1);
2078 u_int32_t addr =
2079 cmd->opcode == O_IP_DST_SET ?
2080 args->f_id.dst_ip :
2081 args->f_id.src_ip;
2082
2083 if (addr < d[0])
2084 break;
2085 addr -= d[0]; /* subtract base */
2086 match = (addr < cmd->arg1) &&
2087 ( d[ 1 + (addr>>5)] &
2088 (1<<(addr & 0x1f)) );
2089 }
2090 break;
2091
2092 case O_IP_DST:
2093 match = (hlen > 0 &&
2094 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
2095 dst_ip.s_addr);
2096 break;
2097
2098 case O_IP_DST_ME:
2099 if (hlen > 0) {
2100 struct ifnet *tif;
2101
2102 INADDR_TO_IFP(dst_ip, tif);
2103 match = (tif != NULL);
2104 }
2105 break;
2106
2107 case O_IP_SRCPORT:
2108 case O_IP_DSTPORT:
2109 /*
2110 * offset == 0 && proto != 0 is enough
2111 * to guarantee that we have an IPv4
2112 * packet with port info.
2113 */
2114 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
2115 && offset == 0) {
2116 u_int16_t x =
2117 (cmd->opcode == O_IP_SRCPORT) ?
2118 src_port : dst_port ;
2119 u_int16_t *p =
2120 ((ipfw_insn_u16 *)cmd)->ports;
2121 int i;
2122
2123 for (i = cmdlen - 1; !match && i>0;
2124 i--, p += 2)
2125 match = (x>=p[0] && x<=p[1]);
2126 }
2127 break;
2128
2129 case O_ICMPTYPE:
2130 match = (offset == 0 && proto==IPPROTO_ICMP &&
2131 icmptype_match(ip, (ipfw_insn_u32 *)cmd) );
2132 break;
2133
2134 case O_IPOPT:
2135 match = (hlen > 0 && ipopts_match(ip, cmd) );
2136 break;
2137
2138 case O_IPVER:
2139 match = (hlen > 0 && cmd->arg1 == ip->ip_v);
2140 break;
2141
2142 case O_IPID:
2143 case O_IPLEN:
2144 case O_IPTTL:
2145 if (hlen > 0) { /* only for IP packets */
2146 uint16_t x;
2147 uint16_t *p;
2148 int i;
2149
2150 if (cmd->opcode == O_IPLEN)
2151 x = ip_len;
2152 else if (cmd->opcode == O_IPTTL)
2153 x = ip->ip_ttl;
2154 else /* must be IPID */
2155 x = ntohs(ip->ip_id);
2156 if (cmdlen == 1) {
2157 match = (cmd->arg1 == x);
2158 break;
2159 }
2160 /* otherwise we have ranges */
2161 p = ((ipfw_insn_u16 *)cmd)->ports;
2162 i = cmdlen - 1;
2163 for (; !match && i>0; i--, p += 2)
2164 match = (x >= p[0] && x <= p[1]);
2165 }
2166 break;
2167
2168 case O_IPPRECEDENCE:
2169 match = (hlen > 0 &&
2170 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2171 break;
2172
2173 case O_IPTOS:
2174 match = (hlen > 0 &&
2175 flags_match(cmd, ip->ip_tos));
2176 break;
2177
2178 case O_TCPFLAGS:
2179 match = (proto == IPPROTO_TCP && offset == 0 &&
2180 flags_match(cmd,
2181 L3HDR(struct tcphdr,ip)->th_flags));
2182 break;
2183
2184 case O_TCPOPTS:
2185 match = (proto == IPPROTO_TCP && offset == 0 &&
2186 tcpopts_match(ip, cmd));
2187 break;
2188
2189 case O_TCPSEQ:
2190 match = (proto == IPPROTO_TCP && offset == 0 &&
2191 ((ipfw_insn_u32 *)cmd)->d[0] ==
2192 L3HDR(struct tcphdr,ip)->th_seq);
2193 break;
2194
2195 case O_TCPACK:
2196 match = (proto == IPPROTO_TCP && offset == 0 &&
2197 ((ipfw_insn_u32 *)cmd)->d[0] ==
2198 L3HDR(struct tcphdr,ip)->th_ack);
2199 break;
2200
2201 case O_TCPWIN:
2202 match = (proto == IPPROTO_TCP && offset == 0 &&
2203 cmd->arg1 ==
2204 L3HDR(struct tcphdr,ip)->th_win);
2205 break;
2206
2207 case O_ESTAB:
2208 /* reject packets which have SYN only */
2209 /* XXX should i also check for TH_ACK ? */
2210 match = (proto == IPPROTO_TCP && offset == 0 &&
2211 (L3HDR(struct tcphdr,ip)->th_flags &
2212 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2213 break;
2214
2215 case O_LOG:
2216 if (fw_verbose)
2217 ipfw_log(f, hlen, args->eh, m, oif);
2218 match = 1;
2219 break;
2220
2221 case O_PROB:
2222 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2223 break;
2224
2225 case O_VERREVPATH:
2226 /* Outgoing packets automatically pass/match */
2227 match = (hlen > 0 && ((oif != NULL) ||
2228 (m->m_pkthdr.rcvif == NULL) ||
2229 verify_path(src_ip, m->m_pkthdr.rcvif)));
2230 break;
2231
2232 case O_VERSRCREACH:
2233 /* Outgoing packets automatically pass/match */
2234 match = (hlen > 0 && ((oif != NULL) ||
2235 verify_path(src_ip, NULL)));
2236 break;
2237
2238 case O_ANTISPOOF:
2239 /* Outgoing packets automatically pass/match */
2240 if (oif == NULL && hlen > 0 &&
2241 in_localaddr(src_ip))
2242 match = verify_path(src_ip,
2243 m->m_pkthdr.rcvif);
2244 else
2245 match = 1;
2246 break;
2247
2248 case O_IPSEC:
2249 #ifdef FAST_IPSEC
2250 match = (m_tag_find(m,
2251 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2252 #endif
2253 #ifdef IPSEC
2254 match = (ipsec_getnhist(m) != 0);
2255 #endif
2256 /* otherwise no match */
2257 break;
2258
2259 /*
2260 * The second set of opcodes represents 'actions',
2261 * i.e. the terminal part of a rule once the packet
2262 * matches all previous patterns.
2263 * Typically there is only one action for each rule,
2264 * and the opcode is stored at the end of the rule
2265 * (but there are exceptions -- see below).
2266 *
2267 * In general, here we set retval and terminate the
2268 * outer loop (would be a 'break 3' in some language,
2269 * but we need to do a 'goto done').
2270 *
2271 * Exceptions:
2272 * O_COUNT and O_SKIPTO actions:
2273 * instead of terminating, we jump to the next rule
2274 * ('goto next_rule', equivalent to a 'break 2'),
2275 * or to the SKIPTO target ('goto again' after
2276 * having set f, cmd and l), respectively.
2277 *
2278 * O_LIMIT and O_KEEP_STATE: these opcodes are
2279 * not real 'actions', and are stored right
2280 * before the 'action' part of the rule.
2281 * These opcodes try to install an entry in the
2282 * state tables; if successful, we continue with
2283 * the next opcode (match=1; break;), otherwise
2284 * the packet * must be dropped
2285 * ('goto done' after setting retval);
2286 *
2287 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2288 * cause a lookup of the state table, and a jump
2289 * to the 'action' part of the parent rule
2290 * ('goto check_body') if an entry is found, or
2291 * (CHECK_STATE only) a jump to the next rule if
2292 * the entry is not found ('goto next_rule').
2293 * The result of the lookup is cached to make
2294 * further instances of these opcodes are
2295 * effectively NOPs.
2296 */
2297 case O_LIMIT:
2298 case O_KEEP_STATE:
2299 if (install_state(f,
2300 (ipfw_insn_limit *)cmd, args)) {
2301 retval = IP_FW_PORT_DENY_FLAG;
2302 goto done; /* error/limit violation */
2303 }
2304 match = 1;
2305 break;
2306
2307 case O_PROBE_STATE:
2308 case O_CHECK_STATE:
2309 /*
2310 * dynamic rules are checked at the first
2311 * keep-state or check-state occurrence,
2312 * with the result being stored in dyn_dir.
2313 * The compiler introduces a PROBE_STATE
2314 * instruction for us when we have a
2315 * KEEP_STATE (because PROBE_STATE needs
2316 * to be run first).
2317 */
2318 if (dyn_dir == MATCH_UNKNOWN &&
2319 (q = lookup_dyn_rule(&args->f_id,
2320 &dyn_dir, proto == IPPROTO_TCP ?
2321 L3HDR(struct tcphdr, ip) : NULL))
2322 != NULL) {
2323 /*
2324 * Found dynamic entry, update stats
2325 * and jump to the 'action' part of
2326 * the parent rule.
2327 */
2328 q->pcnt++;
2329 q->bcnt += pktlen;
2330 f = q->rule;
2331 cmd = ACTION_PTR(f);
2332 l = f->cmd_len - f->act_ofs;
2333 IPFW_DYN_UNLOCK();
2334 goto check_body;
2335 }
2336 /*
2337 * Dynamic entry not found. If CHECK_STATE,
2338 * skip to next rule, if PROBE_STATE just
2339 * ignore and continue with next opcode.
2340 */
2341 if (cmd->opcode == O_CHECK_STATE)
2342 goto next_rule;
2343 match = 1;
2344 break;
2345
2346 case O_ACCEPT:
2347 retval = 0; /* accept */
2348 goto done;
2349
2350 case O_PIPE:
2351 case O_QUEUE:
2352 args->rule = f; /* report matching rule */
2353 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG;
2354 goto done;
2355
2356 case O_DIVERT:
2357 case O_TEE: {
2358 struct divert_tag *dt;
2359
2360 if (args->eh) /* not on layer 2 */
2361 break;
2362 mtag = m_tag_get(PACKET_TAG_DIVERT,
2363 sizeof(struct divert_tag),
2364 M_NOWAIT);
2365 if (mtag == NULL) {
2366 /* XXX statistic */
2367 /* drop packet */
2368 IPFW_UNLOCK(chain);
2369 return IP_FW_PORT_DENY_FLAG;
2370 }
2371 dt = (struct divert_tag *)(mtag+1);
2372 dt->cookie = f->rulenum;
2373 dt->info = (cmd->opcode == O_DIVERT) ?
2374 cmd->arg1 :
2375 cmd->arg1 | IP_FW_PORT_TEE_FLAG;
2376 m_tag_prepend(m, mtag);
2377 retval = dt->info;
2378 goto done;
2379 }
2380
2381 case O_COUNT:
2382 case O_SKIPTO:
2383 f->pcnt++; /* update stats */
2384 f->bcnt += pktlen;
2385 f->timestamp = time_second;
2386 if (cmd->opcode == O_COUNT)
2387 goto next_rule;
2388 /* handle skipto */
2389 if (f->next_rule == NULL)
2390 lookup_next_rule(f);
2391 f = f->next_rule;
2392 goto again;
2393
2394 case O_REJECT:
2395 /*
2396 * Drop the packet and send a reject notice
2397 * if the packet is not ICMP (or is an ICMP
2398 * query), and it is not multicast/broadcast.
2399 */
2400 if (hlen > 0 &&
2401 (proto != IPPROTO_ICMP ||
2402 is_icmp_query(ip)) &&
2403 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2404 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2405 send_reject(args, cmd->arg1,
2406 offset,ip_len);
2407 m = args->m;
2408 }
2409 /* FALLTHROUGH */
2410 case O_DENY:
2411 retval = IP_FW_PORT_DENY_FLAG;
2412 goto done;
2413
2414 case O_FORWARD_IP:
2415 if (args->eh) /* not valid on layer2 pkts */
2416 break;
2417 if (!q || dyn_dir == MATCH_FORWARD)
2418 args->next_hop =
2419 &((ipfw_insn_sa *)cmd)->sa;
2420 retval = 0;
2421 goto done;
2422
2423 default:
2424 panic("-- unknown opcode %d\n", cmd->opcode);
2425 } /* end of switch() on opcodes */
2426
2427 if (cmd->len & F_NOT)
2428 match = !match;
2429
2430 if (match) {
2431 if (cmd->len & F_OR)
2432 skip_or = 1;
2433 } else {
2434 if (!(cmd->len & F_OR)) /* not an OR block, */
2435 break; /* try next rule */
2436 }
2437
2438 } /* end of inner for, scan opcodes */
2439
2440 next_rule:; /* try next rule */
2441
2442 } /* end of outer for, scan rules */
2443 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2444 IPFW_UNLOCK(chain);
2445 return(IP_FW_PORT_DENY_FLAG);
2446
2447 done:
2448 /* Update statistics */
2449 f->pcnt++;
2450 f->bcnt += pktlen;
2451 f->timestamp = time_second;
2452 IPFW_UNLOCK(chain);
2453 return retval;
2454
2455 pullup_failed:
2456 if (fw_verbose)
2457 printf("ipfw: pullup failed\n");
2458 return(IP_FW_PORT_DENY_FLAG);
2459 }
2460
2461 /*
2462 * When a rule is added/deleted, clear the next_rule pointers in all rules.
2463 * These will be reconstructed on the fly as packets are matched.
2464 */
2465 static void
2466 flush_rule_ptrs(struct ip_fw_chain *chain)
2467 {
2468 struct ip_fw *rule;
2469
2470 IPFW_LOCK_ASSERT(chain);
2471
2472 for (rule = chain->rules; rule; rule = rule->next)
2473 rule->next_rule = NULL;
2474 }
2475
2476 /*
2477 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given
2478 * pipe/queue, or to all of them (match == NULL).
2479 */
2480 void
2481 flush_pipe_ptrs(struct dn_flow_set *match)
2482 {
2483 struct ip_fw *rule;
2484
2485 IPFW_LOCK(&layer3_chain);
2486 for (rule = layer3_chain.rules; rule; rule = rule->next) {
2487 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule);
2488
2489 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE)
2490 continue;
2491 /*
2492 * XXX Use bcmp/bzero to handle pipe_ptr to overcome
2493 * possible alignment problems on 64-bit architectures.
2494 * This code is seldom used so we do not worry too
2495 * much about efficiency.
2496 */
2497 if (match == NULL ||
2498 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) )
2499 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr));
2500 }
2501 IPFW_UNLOCK(&layer3_chain);
2502 }
2503
2504 /*
2505 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
2506 * possibly create a rule number and add the rule to the list.
2507 * Update the rule_number in the input struct so the caller knows it as well.
2508 */
2509 static int
2510 add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule)
2511 {
2512 struct ip_fw *rule, *f, *prev;
2513 int l = RULESIZE(input_rule);
2514
2515 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE)
2516 return (EINVAL);
2517
2518 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO);
2519 if (rule == NULL)
2520 return (ENOSPC);
2521
2522 bcopy(input_rule, rule, l);
2523
2524 rule->next = NULL;
2525 rule->next_rule = NULL;
2526
2527 rule->pcnt = 0;
2528 rule->bcnt = 0;
2529 rule->timestamp = 0;
2530
2531 IPFW_LOCK(chain);
2532
2533 if (chain->rules == NULL) { /* default rule */
2534 chain->rules = rule;
2535 goto done;
2536 }
2537
2538 /*
2539 * If rulenum is 0, find highest numbered rule before the
2540 * default rule, and add autoinc_step
2541 */
2542 if (autoinc_step < 1)
2543 autoinc_step = 1;
2544 else if (autoinc_step > 1000)
2545 autoinc_step = 1000;
2546 if (rule->rulenum == 0) {
2547 /*
2548 * locate the highest numbered rule before default
2549 */
2550 for (f = chain->rules; f; f = f->next) {
2551 if (f->rulenum == IPFW_DEFAULT_RULE)
2552 break;
2553 rule->rulenum = f->rulenum;
2554 }
2555 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
2556 rule->rulenum += autoinc_step;
2557 input_rule->rulenum = rule->rulenum;
2558 }
2559
2560 /*
2561 * Now insert the new rule in the right place in the sorted list.
2562 */
2563 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) {
2564 if (f->rulenum > rule->rulenum) { /* found the location */
2565 if (prev) {
2566 rule->next = f;
2567 prev->next = rule;
2568 } else { /* head insert */
2569 rule->next = chain->rules;
2570 chain->rules = rule;
2571 }
2572 break;
2573 }
2574 }
2575 flush_rule_ptrs(chain);
2576 done:
2577 static_count++;
2578 static_len += l;
2579 IPFW_UNLOCK(chain);
2580 DEB(printf("ipfw: installed rule %d, static count now %d\n",
2581 rule->rulenum, static_count);)
2582 return (0);
2583 }
2584
2585 /**
2586 * Remove a static rule (including derived * dynamic rules)
2587 * and place it on the ``reap list'' for later reclamation.
2588 * The caller is in charge of clearing rule pointers to avoid
2589 * dangling pointers.
2590 * @return a pointer to the next entry.
2591 * Arguments are not checked, so they better be correct.
2592 */
2593 static struct ip_fw *
2594 remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev)
2595 {
2596 struct ip_fw *n;
2597 int l = RULESIZE(rule);
2598
2599 IPFW_LOCK_ASSERT(chain);
2600
2601 n = rule->next;
2602 IPFW_DYN_LOCK();
2603 remove_dyn_rule(rule, NULL /* force removal */);
2604 IPFW_DYN_UNLOCK();
2605 if (prev == NULL)
2606 chain->rules = n;
2607 else
2608 prev->next = n;
2609 static_count--;
2610 static_len -= l;
2611
2612 rule->next = chain->reap;
2613 chain->reap = rule;
2614
2615 return n;
2616 }
2617
2618 /**
2619 * Reclaim storage associated with a list of rules. This is
2620 * typically the list created using remove_rule.
2621 */
2622 static void
2623 reap_rules(struct ip_fw *head)
2624 {
2625 struct ip_fw *rule;
2626
2627 while ((rule = head) != NULL) {
2628 head = head->next;
2629 if (DUMMYNET_LOADED)
2630 ip_dn_ruledel_ptr(rule);
2631 free(rule, M_IPFW);
2632 }
2633 }
2634
2635 /*
2636 * Remove all rules from a chain (except rules in set RESVD_SET
2637 * unless kill_default = 1). The caller is responsible for
2638 * reclaiming storage for the rules left in chain->reap.
2639 */
2640 static void
2641 free_chain(struct ip_fw_chain *chain, int kill_default)
2642 {
2643 struct ip_fw *prev, *rule;
2644
2645 IPFW_LOCK_ASSERT(chain);
2646
2647 flush_rule_ptrs(chain); /* more efficient to do outside the loop */
2648 for (prev = NULL, rule = chain->rules; rule ; )
2649 if (kill_default || rule->set != RESVD_SET)
2650 rule = remove_rule(chain, rule, prev);
2651 else {
2652 prev = rule;
2653 rule = rule->next;
2654 }
2655 }
2656
2657 /**
2658 * Remove all rules with given number, and also do set manipulation.
2659 * Assumes chain != NULL && *chain != NULL.
2660 *
2661 * The argument is an u_int32_t. The low 16 bit are the rule or set number,
2662 * the next 8 bits are the new set, the top 8 bits are the command:
2663 *
2664 * 0 delete rules with given number
2665 * 1 delete rules with given set number
2666 * 2 move rules with given number to new set
2667 * 3 move rules with given set number to new set
2668 * 4 swap sets with given numbers
2669 */
2670 static int
2671 del_entry(struct ip_fw_chain *chain, u_int32_t arg)
2672 {
2673 struct ip_fw *prev = NULL, *rule;
2674 u_int16_t rulenum; /* rule or old_set */
2675 u_int8_t cmd, new_set;
2676
2677 rulenum = arg & 0xffff;
2678 cmd = (arg >> 24) & 0xff;
2679 new_set = (arg >> 16) & 0xff;
2680
2681 if (cmd > 4)
2682 return EINVAL;
2683 if (new_set > RESVD_SET)
2684 return EINVAL;
2685 if (cmd == 0 || cmd == 2) {
2686 if (rulenum >= IPFW_DEFAULT_RULE)
2687 return EINVAL;
2688 } else {
2689 if (rulenum > RESVD_SET) /* old_set */
2690 return EINVAL;
2691 }
2692
2693 IPFW_LOCK(chain);
2694 rule = chain->rules;
2695 chain->reap = NULL;
2696 switch (cmd) {
2697 case 0: /* delete rules with given number */
2698 /*
2699 * locate first rule to delete
2700 */
2701 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
2702 ;
2703 if (rule->rulenum != rulenum) {
2704 IPFW_UNLOCK(chain);
2705 return EINVAL;
2706 }
2707
2708 /*
2709 * flush pointers outside the loop, then delete all matching
2710 * rules. prev remains the same throughout the cycle.
2711 */
2712 flush_rule_ptrs(chain);
2713 while (rule->rulenum == rulenum)
2714 rule = remove_rule(chain, rule, prev);
2715 break;
2716
2717 case 1: /* delete all rules with given set number */
2718 flush_rule_ptrs(chain);
2719 rule = chain->rules;
2720 while (rule->rulenum < IPFW_DEFAULT_RULE)
2721 if (rule->set == rulenum)
2722 rule = remove_rule(chain, rule, prev);
2723 else {
2724 prev = rule;
2725 rule = rule->next;
2726 }
2727 break;
2728
2729 case 2: /* move rules with given number to new set */
2730 rule = chain->rules;
2731 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
2732 if (rule->rulenum == rulenum)
2733 rule->set = new_set;
2734 break;
2735
2736 case 3: /* move rules with given set number to new set */
2737 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
2738 if (rule->set == rulenum)
2739 rule->set = new_set;
2740 break;
2741
2742 case 4: /* swap two sets */
2743 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
2744 if (rule->set == rulenum)
2745 rule->set = new_set;
2746 else if (rule->set == new_set)
2747 rule->set = rulenum;
2748 break;
2749 }
2750 /*
2751 * Look for rules to reclaim. We grab the list before
2752 * releasing the lock then reclaim them w/o the lock to
2753 * avoid a LOR with dummynet.
2754 */
2755 rule = chain->reap;
2756 chain->reap = NULL;
2757 IPFW_UNLOCK(chain);
2758 if (rule)
2759 reap_rules(rule);
2760 return 0;
2761 }
2762
2763 /*
2764 * Clear counters for a specific rule.
2765 * The enclosing "table" is assumed locked.
2766 */
2767 static void
2768 clear_counters(struct ip_fw *rule, int log_only)
2769 {
2770 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
2771
2772 if (log_only == 0) {
2773 rule->bcnt = rule->pcnt = 0;
2774 rule->timestamp = 0;
2775 }
2776 if (l->o.opcode == O_LOG)
2777 l->log_left = l->max_log;
2778 }
2779
2780 /**
2781 * Reset some or all counters on firewall rules.
2782 * @arg frwl is null to clear all entries, or contains a specific
2783 * rule number.
2784 * @arg log_only is 1 if we only want to reset logs, zero otherwise.
2785 */
2786 static int
2787 zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only)
2788 {
2789 struct ip_fw *rule;
2790 char *msg;
2791
2792 IPFW_LOCK(chain);
2793 if (rulenum == 0) {
2794 norule_counter = 0;
2795 for (rule = chain->rules; rule; rule = rule->next)
2796 clear_counters(rule, log_only);
2797 msg = log_only ? "ipfw: All logging counts reset.\n" :
2798 "ipfw: Accounting cleared.\n";
2799 } else {
2800 int cleared = 0;
2801 /*
2802 * We can have multiple rules with the same number, so we
2803 * need to clear them all.
2804 */
2805 for (rule = chain->rules; rule; rule = rule->next)
2806 if (rule->rulenum == rulenum) {
2807 while (rule && rule->rulenum == rulenum) {
2808 clear_counters(rule, log_only);
2809 rule = rule->next;
2810 }
2811 cleared = 1;
2812 break;
2813 }
2814 if (!cleared) { /* we did not find any matching rules */
2815 IPFW_UNLOCK(chain);
2816 return (EINVAL);
2817 }
2818 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
2819 "ipfw: Entry %d cleared.\n";
2820 }
2821 IPFW_UNLOCK(chain);
2822
2823 if (fw_verbose)
2824 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum);
2825 return (0);
2826 }
2827
2828 /*
2829 * Check validity of the structure before insert.
2830 * Fortunately rules are simple, so this mostly need to check rule sizes.
2831 */
2832 static int
2833 check_ipfw_struct(struct ip_fw *rule, int size)
2834 {
2835 int l, cmdlen = 0;
2836 int have_action=0;
2837 ipfw_insn *cmd;
2838
2839 if (size < sizeof(*rule)) {
2840 printf("ipfw: rule too short\n");
2841 return (EINVAL);
2842 }
2843 /* first, check for valid size */
2844 l = RULESIZE(rule);
2845 if (l != size) {
2846 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
2847 return (EINVAL);
2848 }
2849 if (rule->act_ofs >= rule->cmd_len) {
2850 printf("ipfw: bogus action offset (%u > %u)\n",
2851 rule->act_ofs, rule->cmd_len - 1);
2852 return (EINVAL);
2853 }
2854 /*
2855 * Now go for the individual checks. Very simple ones, basically only
2856 * instruction sizes.
2857 */
2858 for (l = rule->cmd_len, cmd = rule->cmd ;
2859 l > 0 ; l -= cmdlen, cmd += cmdlen) {
2860 cmdlen = F_LEN(cmd);
2861 if (cmdlen > l) {
2862 printf("ipfw: opcode %d size truncated\n",
2863 cmd->opcode);
2864 return EINVAL;
2865 }
2866 DEB(printf("ipfw: opcode %d\n", cmd->opcode);)
2867 switch (cmd->opcode) {
2868 case O_PROBE_STATE:
2869 case O_KEEP_STATE:
2870 case O_PROTO:
2871 case O_IP_SRC_ME:
2872 case O_IP_DST_ME:
2873 case O_LAYER2:
2874 case O_IN:
2875 case O_FRAG:
2876 case O_IPOPT:
2877 case O_IPTOS:
2878 case O_IPPRECEDENCE:
2879 case O_IPVER:
2880 case O_TCPWIN:
2881 case O_TCPFLAGS:
2882 case O_TCPOPTS:
2883 case O_ESTAB:
2884 case O_VERREVPATH:
2885 case O_VERSRCREACH:
2886 case O_ANTISPOOF:
2887 case O_IPSEC:
2888 if (cmdlen != F_INSN_SIZE(ipfw_insn))
2889 goto bad_size;
2890 break;
2891
2892 case O_UID:
2893 case O_GID:
2894 case O_JAIL:
2895 case O_IP_SRC:
2896 case O_IP_DST:
2897 case O_TCPSEQ:
2898 case O_TCPACK:
2899 case O_PROB:
2900 case O_ICMPTYPE:
2901 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
2902 goto bad_size;
2903 break;
2904
2905 case O_LIMIT:
2906 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
2907 goto bad_size;
2908 break;
2909
2910 case O_LOG:
2911 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
2912 goto bad_size;
2913
2914 ((ipfw_insn_log *)cmd)->log_left =
2915 ((ipfw_insn_log *)cmd)->max_log;
2916
2917 break;
2918
2919 case O_IP_SRC_MASK:
2920 case O_IP_DST_MASK:
2921 /* only odd command lengths */
2922 if ( !(cmdlen & 1) || cmdlen > 31)
2923 goto bad_size;
2924 break;
2925
2926 case O_IP_SRC_SET:
2927 case O_IP_DST_SET:
2928 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
2929 printf("ipfw: invalid set size %d\n",
2930 cmd->arg1);
2931 return EINVAL;
2932 }
2933 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
2934 (cmd->arg1+31)/32 )
2935 goto bad_size;
2936 break;
2937
2938 case O_IP_SRC_LOOKUP:
2939 case O_IP_DST_LOOKUP:
2940 if (cmd->arg1 >= IPFW_TABLES_MAX) {
2941 printf("ipfw: invalid table number %d\n",
2942 cmd->arg1);
2943 return (EINVAL);
2944 }
2945 if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
2946 cmdlen != F_INSN_SIZE(ipfw_insn_u32))
2947 goto bad_size;
2948 break;
2949
2950 case O_MACADDR2:
2951 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
2952 goto bad_size;
2953 break;
2954
2955 case O_NOP:
2956 case O_IPID:
2957 case O_IPTTL:
2958 case O_IPLEN:
2959 if (cmdlen < 1 || cmdlen > 31)
2960 goto bad_size;
2961 break;
2962
2963 case O_MAC_TYPE:
2964 case O_IP_SRCPORT:
2965 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
2966 if (cmdlen < 2 || cmdlen > 31)
2967 goto bad_size;
2968 break;
2969
2970 case O_RECV:
2971 case O_XMIT:
2972 case O_VIA:
2973 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
2974 goto bad_size;
2975 break;
2976
2977 case O_PIPE:
2978 case O_QUEUE:
2979 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe))
2980 goto bad_size;
2981 goto check_action;
2982
2983 case O_FORWARD_IP:
2984 #ifdef IPFIREWALL_FORWARD
2985 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
2986 goto bad_size;
2987 goto check_action;
2988 #else
2989 return EINVAL;
2990 #endif
2991
2992 case O_DIVERT:
2993 case O_TEE:
2994 #ifndef IPDIVERT
2995 return EINVAL;
2996 #endif
2997 case O_FORWARD_MAC: /* XXX not implemented yet */
2998 case O_CHECK_STATE:
2999 case O_COUNT:
3000 case O_ACCEPT:
3001 case O_DENY:
3002 case O_REJECT:
3003 case O_SKIPTO:
3004 if (cmdlen != F_INSN_SIZE(ipfw_insn))
3005 goto bad_size;
3006 check_action:
3007 if (have_action) {
3008 printf("ipfw: opcode %d, multiple actions"
3009 " not allowed\n",
3010 cmd->opcode);
3011 return EINVAL;
3012 }
3013 have_action = 1;
3014 if (l != cmdlen) {
3015 printf("ipfw: opcode %d, action must be"
3016 " last opcode\n",
3017 cmd->opcode);
3018 return EINVAL;
3019 }
3020 break;
3021 default:
3022 printf("ipfw: opcode %d, unknown opcode\n",
3023 cmd->opcode);
3024 return EINVAL;
3025 }
3026 }
3027 if (have_action == 0) {
3028 printf("ipfw: missing action\n");
3029 return EINVAL;
3030 }
3031 return 0;
3032
3033 bad_size:
3034 printf("ipfw: opcode %d size %d wrong\n",
3035 cmd->opcode, cmdlen);
3036 return EINVAL;
3037 }
3038
3039 /*
3040 * Copy the static and dynamic rules to the supplied buffer
3041 * and return the amount of space actually used.
3042 */
3043 static size_t
3044 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
3045 {
3046 char *bp = buf;
3047 char *ep = bp + space;
3048 struct ip_fw *rule;
3049 int i;
3050
3051 /* XXX this can take a long time and locking will block packet flow */
3052 IPFW_LOCK(chain);
3053 for (rule = chain->rules; rule ; rule = rule->next) {
3054 /*
3055 * Verify the entry fits in the buffer in case the
3056 * rules changed between calculating buffer space and
3057 * now. This would be better done using a generation
3058 * number but should suffice for now.
3059 */
3060 i = RULESIZE(rule);
3061 if (bp + i <= ep) {
3062 bcopy(rule, bp, i);
3063 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule),
3064 sizeof(set_disable));
3065 bp += i;
3066 }
3067 }
3068 IPFW_UNLOCK(chain);
3069 if (ipfw_dyn_v) {
3070 ipfw_dyn_rule *p, *last = NULL;
3071
3072 IPFW_DYN_LOCK();
3073 for (i = 0 ; i < curr_dyn_buckets; i++)
3074 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) {
3075 if (bp + sizeof *p <= ep) {
3076 ipfw_dyn_rule *dst =
3077 (ipfw_dyn_rule *)bp;
3078 bcopy(p, dst, sizeof *p);
3079 bcopy(&(p->rule->rulenum), &(dst->rule),
3080 sizeof(p->rule->rulenum));
3081 /*
3082 * store a non-null value in "next".
3083 * The userland code will interpret a
3084 * NULL here as a marker
3085 * for the last dynamic rule.
3086 */
3087 bcopy(&dst, &dst->next, sizeof(dst));
3088 last = dst;
3089 dst->expire =
3090 TIME_LEQ(dst->expire, time_second) ?
3091 0 : dst->expire - time_second ;
3092 bp += sizeof(ipfw_dyn_rule);
3093 }
3094 }
3095 IPFW_DYN_UNLOCK();
3096 if (last != NULL) /* mark last dynamic rule */
3097 bzero(&last->next, sizeof(last));
3098 }
3099 return (bp - (char *)buf);
3100 }
3101
3102
3103 /**
3104 * {set|get}sockopt parser.
3105 */
3106 static int
3107 ipfw_ctl(struct sockopt *sopt)
3108 {
3109 #define RULE_MAXSIZE (256*sizeof(u_int32_t))
3110 int error, rule_num;
3111 size_t size;
3112 struct ip_fw *buf, *rule;
3113 u_int32_t rulenum[2];
3114
3115 error = suser(sopt->sopt_td);
3116 if (error)
3117 return (error);
3118
3119 /*
3120 * Disallow modifications in really-really secure mode, but still allow
3121 * the logging counters to be reset.
3122 */
3123 if (sopt->sopt_name == IP_FW_ADD ||
3124 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
3125 #if __FreeBSD_version >= 500034
3126 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
3127 if (error)
3128 return (error);
3129 #else /* FreeBSD 4.x */
3130 if (securelevel >= 3)
3131 return (EPERM);
3132 #endif
3133 }
3134
3135 error = 0;
3136
3137 switch (sopt->sopt_name) {
3138 case IP_FW_GET:
3139 /*
3140 * pass up a copy of the current rules. Static rules
3141 * come first (the last of which has number IPFW_DEFAULT_RULE),
3142 * followed by a possibly empty list of dynamic rule.
3143 * The last dynamic rule has NULL in the "next" field.
3144 *
3145 * Note that the calculated size is used to bound the
3146 * amount of data returned to the user. The rule set may
3147 * change between calculating the size and returning the
3148 * data in which case we'll just return what fits.
3149 */
3150 size = static_len; /* size of static rules */
3151 if (ipfw_dyn_v) /* add size of dyn.rules */
3152 size += (dyn_count * sizeof(ipfw_dyn_rule));
3153
3154 /*
3155 * XXX todo: if the user passes a short length just to know
3156 * how much room is needed, do not bother filling up the
3157 * buffer, just jump to the sooptcopyout.
3158 */
3159 buf = malloc(size, M_TEMP, M_WAITOK);
3160 error = sooptcopyout(sopt, buf,
3161 ipfw_getrules(&layer3_chain, buf, size));
3162 free(buf, M_TEMP);
3163 break;
3164
3165 case IP_FW_FLUSH:
3166 /*
3167 * Normally we cannot release the lock on each iteration.
3168 * We could do it here only because we start from the head all
3169 * the times so there is no risk of missing some entries.
3170 * On the other hand, the risk is that we end up with
3171 * a very inconsistent ruleset, so better keep the lock
3172 * around the whole cycle.
3173 *
3174 * XXX this code can be improved by resetting the head of
3175 * the list to point to the default rule, and then freeing
3176 * the old list without the need for a lock.
3177 */
3178
3179 IPFW_LOCK(&layer3_chain);
3180 layer3_chain.reap = NULL;
3181 free_chain(&layer3_chain, 0 /* keep default rule */);
3182 rule = layer3_chain.reap, layer3_chain.reap = NULL;
3183 IPFW_UNLOCK(&layer3_chain);
3184 if (layer3_chain.reap != NULL)
3185 reap_rules(rule);
3186 break;
3187
3188 case IP_FW_ADD:
3189 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
3190 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
3191 sizeof(struct ip_fw) );
3192 if (error == 0)
3193 error = check_ipfw_struct(rule, sopt->sopt_valsize);
3194 if (error == 0) {
3195 error = add_rule(&layer3_chain, rule);
3196 size = RULESIZE(rule);
3197 if (!error && sopt->sopt_dir == SOPT_GET)
3198 error = sooptcopyout(sopt, rule, size);
3199 }
3200 free(rule, M_TEMP);
3201 break;
3202
3203 case IP_FW_DEL:
3204 /*
3205 * IP_FW_DEL is used for deleting single rules or sets,
3206 * and (ab)used to atomically manipulate sets. Argument size
3207 * is used to distinguish between the two:
3208 * sizeof(u_int32_t)
3209 * delete single rule or set of rules,
3210 * or reassign rules (or sets) to a different set.
3211 * 2*sizeof(u_int32_t)
3212 * atomic disable/enable sets.
3213 * first u_int32_t contains sets to be disabled,
3214 * second u_int32_t contains sets to be enabled.
3215 */
3216 error = sooptcopyin(sopt, rulenum,
3217 2*sizeof(u_int32_t), sizeof(u_int32_t));
3218 if (error)
3219 break;
3220 size = sopt->sopt_valsize;
3221 if (size == sizeof(u_int32_t)) /* delete or reassign */
3222 error = del_entry(&layer3_chain, rulenum[0]);
3223 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */
3224 set_disable =
3225 (set_disable | rulenum[0]) & ~rulenum[1] &
3226 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
3227 else
3228 error = EINVAL;
3229 break;
3230
3231 case IP_FW_ZERO:
3232 case IP_FW_RESETLOG: /* argument is an int, the rule number */
3233 rule_num = 0;
3234 if (sopt->sopt_val != 0) {
3235 error = sooptcopyin(sopt, &rule_num,
3236 sizeof(int), sizeof(int));
3237 if (error)
3238 break;
3239 }
3240 error = zero_entry(&layer3_chain, rule_num,
3241 sopt->sopt_name == IP_FW_RESETLOG);
3242 break;
3243
3244 case IP_FW_TABLE_ADD:
3245 {
3246 ipfw_table_entry ent;
3247
3248 error = sooptcopyin(sopt, &ent,
3249 sizeof(ent), sizeof(ent));
3250 if (error)
3251 break;
3252 error = add_table_entry(ent.tbl, ent.addr,
3253 ent.masklen, ent.value);
3254 }
3255 break;
3256
3257 case IP_FW_TABLE_DEL:
3258 {
3259 ipfw_table_entry ent;
3260
3261 error = sooptcopyin(sopt, &ent,
3262 sizeof(ent), sizeof(ent));
3263 if (error)
3264 break;
3265 error = del_table_entry(ent.tbl, ent.addr, ent.masklen);
3266 }
3267 break;
3268
3269 case IP_FW_TABLE_FLUSH:
3270 {
3271 u_int16_t tbl;
3272
3273 error = sooptcopyin(sopt, &tbl,
3274 sizeof(tbl), sizeof(tbl));
3275 if (error)
3276 break;
3277 error = flush_table(tbl);
3278 }
3279 break;
3280
3281 case IP_FW_TABLE_GETSIZE:
3282 {
3283 u_int32_t tbl, cnt;
3284
3285 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
3286 sizeof(tbl))))
3287 break;
3288 if ((error = count_table(tbl, &cnt)))
3289 break;
3290 error = sooptcopyout(sopt, &cnt, sizeof(cnt));
3291 }
3292 break;
3293
3294 case IP_FW_TABLE_LIST:
3295 {
3296 ipfw_table *tbl;
3297
3298 if (sopt->sopt_valsize < sizeof(*tbl)) {
3299 error = EINVAL;
3300 break;
3301 }
3302 size = sopt->sopt_valsize;
3303 tbl = malloc(size, M_TEMP, M_WAITOK);
3304 if (tbl == NULL) {
3305 error = ENOMEM;
3306 break;
3307 }
3308 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
3309 if (error) {
3310 free(tbl, M_TEMP);
3311 break;
3312 }
3313 tbl->size = (size - sizeof(*tbl)) /
3314 sizeof(ipfw_table_entry);
3315 error = dump_table(tbl);
3316 if (error) {
3317 free(tbl, M_TEMP);
3318 break;
3319 }
3320 error = sooptcopyout(sopt, tbl, size);
3321 free(tbl, M_TEMP);
3322 }
3323 break;
3324
3325 default:
3326 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
3327 error = EINVAL;
3328 }
3329
3330 return (error);
3331 #undef RULE_MAXSIZE
3332 }
3333
3334 /**
3335 * dummynet needs a reference to the default rule, because rules can be
3336 * deleted while packets hold a reference to them. When this happens,
3337 * dummynet changes the reference to the default rule (it could well be a
3338 * NULL pointer, but this way we do not need to check for the special
3339 * case, plus here he have info on the default behaviour).
3340 */
3341 struct ip_fw *ip_fw_default_rule;
3342
3343 /*
3344 * This procedure is only used to handle keepalives. It is invoked
3345 * every dyn_keepalive_period
3346 */
3347 static void
3348 ipfw_tick(void * __unused unused)
3349 {
3350 int i;
3351 ipfw_dyn_rule *q;
3352
3353 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
3354 goto done;
3355
3356 IPFW_DYN_LOCK();
3357 for (i = 0 ; i < curr_dyn_buckets ; i++) {
3358 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
3359 if (q->dyn_type == O_LIMIT_PARENT)
3360 continue;
3361 if (q->id.proto != IPPROTO_TCP)
3362 continue;
3363 if ( (q->state & BOTH_SYN) != BOTH_SYN)
3364 continue;
3365 if (TIME_LEQ( time_second+dyn_keepalive_interval,
3366 q->expire))
3367 continue; /* too early */
3368 if (TIME_LEQ(q->expire, time_second))
3369 continue; /* too late, rule expired */
3370
3371 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN);
3372 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0);
3373 }
3374 }
3375 IPFW_DYN_UNLOCK();
3376 done:
3377 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL);
3378 }
3379
3380 int
3381 ipfw_init(void)
3382 {
3383 struct ip_fw default_rule;
3384 int error;
3385
3386 layer3_chain.rules = NULL;
3387 IPFW_LOCK_INIT(&layer3_chain);
3388 IPFW_DYN_LOCK_INIT();
3389 callout_init(&ipfw_timeout, debug_mpsafenet ? CALLOUT_MPSAFE : 0);
3390
3391 bzero(&default_rule, sizeof default_rule);
3392
3393 default_rule.act_ofs = 0;
3394 default_rule.rulenum = IPFW_DEFAULT_RULE;
3395 default_rule.cmd_len = 1;
3396 default_rule.set = RESVD_SET;
3397
3398 default_rule.cmd[0].len = 1;
3399 default_rule.cmd[0].opcode =
3400 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
3401 1 ? O_ACCEPT :
3402 #endif
3403 O_DENY;
3404
3405 error = add_rule(&layer3_chain, &default_rule);
3406 if (error != 0) {
3407 printf("ipfw2: error %u initializing default rule "
3408 "(support disabled)\n", error);
3409 IPFW_DYN_LOCK_DESTROY();
3410 IPFW_LOCK_DESTROY(&layer3_chain);
3411 return (error);
3412 }
3413
3414 ip_fw_default_rule = layer3_chain.rules;
3415 printf("ipfw2 initialized, divert %s, "
3416 "rule-based forwarding "
3417 #ifdef IPFIREWALL_FORWARD
3418 "enabled, "
3419 #else
3420 "disabled, "
3421 #endif
3422 "default to %s, logging ",
3423 #ifdef IPDIVERT
3424 "enabled",
3425 #else
3426 "disabled",
3427 #endif
3428 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny");
3429
3430 #ifdef IPFIREWALL_VERBOSE
3431 fw_verbose = 1;
3432 #endif
3433 #ifdef IPFIREWALL_VERBOSE_LIMIT
3434 verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3435 #endif
3436 if (fw_verbose == 0)
3437 printf("disabled\n");
3438 else if (verbose_limit == 0)
3439 printf("unlimited\n");
3440 else
3441 printf("limited to %d packets/entry by default\n",
3442 verbose_limit);
3443
3444 init_tables();
3445 ip_fw_ctl_ptr = ipfw_ctl;
3446 ip_fw_chk_ptr = ipfw_chk;
3447 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL);
3448
3449 return (0);
3450 }
3451
3452 void
3453 ipfw_destroy(void)
3454 {
3455 struct ip_fw *reap;
3456
3457 ip_fw_chk_ptr = NULL;
3458 ip_fw_ctl_ptr = NULL;
3459 callout_drain(&ipfw_timeout);
3460 IPFW_LOCK(&layer3_chain);
3461 layer3_chain.reap = NULL;
3462 free_chain(&layer3_chain, 1 /* kill default rule */);
3463 reap = layer3_chain.reap, layer3_chain.reap = NULL;
3464 IPFW_UNLOCK(&layer3_chain);
3465 if (reap != NULL)
3466 reap_rules(reap);
3467 flush_tables();
3468 IPFW_DYN_LOCK_DESTROY();
3469 IPFW_LOCK_DESTROY(&layer3_chain);
3470 printf("IP firewall unloaded\n");
3471 }
3472
3473 #endif /* IPFW2 */
Cache object: f8aa02d5cac5f67513394f3c5ba2eb16
|